CN101115051B - Audio signal processing method, system and audio signal transmitting/receiving device - Google Patents

Abstract

The invention discloses an audio signal processing method, system, and audio signal transceiver device. At the audio signal sending end, the residual signal and the masking threshold signal obtained by processing the audio signal are coded into multiple descriptions of the residual signal and multiple descriptions of the masking threshold signal. , and then respectively combine the residual signal descriptions of each channel with one of the multi-channel masking threshold signal descriptions to generate a multi-channel description that contains both the residual signal and the masking threshold signal; at the audio signal receiving end, all the descriptions received The residual signal description and masking threshold signal description contained in each description are split, and then all the residual signal descriptions generated after splitting are decoded into one residual signal, and all the masking threshold signal descriptions generated after splitting are decoded as Mask the threshold signal all the way. The invention can effectively improve the audio quality in the communication process and improve user satisfaction.

Description

Audio signal processing method, system and audio signal transceiving device

technical field

The present invention relates to the communication field, in particular to an audio signal processing method, system, and audio signal transceiving device.

Background technique

With the rapid development of communication technology, the channel bandwidth and transmission rate have been greatly improved, and the processing capabilities of network equipment and communication terminals have also been continuously enhanced; at the same time, various audio coding algorithms have been considerably improved in coding quality and coding efficiency. This makes real-time high-quality audio services rapidly merged into various modern communication systems. However, in actual operation, due to reasons such as network congestion, channel interference and noise, communication networks based on packet switching inevitably face the problems of packet loss and long delay, which leads The quality of audio signals transmitted by mobile communication systems will undoubtedly be severely affected by packet loss and delay.

At present, there are generally two audio signal processing methods for reducing audio quality degradation caused by packet loss, and the two methods will be introduced respectively below.

The first audio signal processing method is: divide the information sources into levels with different priorities according to certain criteria, and then optimize each layer according to the characteristics of the communication channel; Data packaging processing is performed at different levels to form a layered multi-description coded bit stream, including the basic layer multi-description coded bit stream and the enhancement layer multi-description coded bit stream; finally, the above-mentioned bit stream formed by the data packaging process is sent to the receiving end .

When receiving the base layer multi-description coded bit stream, the receiving end can recover the basic information from the source; Small distortions are better at recovering information from the source.

The purpose of the first audio signal processing method is to enable the receiver to recover the information from the source with as little distortion as possible, but its operation method will inevitably introduce the following problems:

1. Only when receiving the base layer multi-description coded bit stream from the source, the receiving end can perform normal decoding; once the base layer multi-description coded bit stream loses packets, the receiving end will not be able to normally perform the decoding process;

2. When the receiver only receives the enhancement layer multi-description coded bit stream, it can only recover part of the enhancement layer information from the source, and cannot reconstruct the main information from the source;

3. In order to enable the receiving end to receive the base layer multi-description coded bit stream normally, and to prevent bit errors or packet loss, it is usually necessary to perform forward error correction or packet loss repetition on the base layer multi-description coded bit stream during transmission. transmission and other processing; this will significantly reduce communication efficiency and increase communication delay;

4. Since the source is layered according to priority, the generated layered coded bit streams will also have different priorities, and the correct decoding of low priority layered coded bit streams must be layered with high priority The correct decoding of the coded bit stream is based on; when there is a problem in the decoding process of the high-priority layered coded bit stream, the correct decoding of the low-priority layered coded bit stream is impossible.

To sum up, since the first audio signal processing method stratifies the source according to the priority, the reliability of improving the audio quality in the communication process is low, and it cannot effectively reduce the communication loss caused by packet loss and delay. The problem of audio quality degradation during the process; and, audio quality degradation will significantly reduce user satisfaction.

The second audio signal processing method is: use the commonly used preprocessor controlled by the psychoacoustic model to process the audio signal to obtain the residual signal that has been removed from the time-domain whitening that is irrelevant to the auditory sense, and the obtained residual signal Perform multi-description coding to obtain two or more description codes; then perform distortion-free entropy coding on the obtained description codes to remove the redundancy of the information source, and finally send the coded bit stream generated after distortion-free entropy coding to the transmission channel .

The second audio signal processing method seeks to improve audio signal quality with psychoacoustic models, but it operates in such a way that it inevitably introduces the following problems:

1. The second audio signal processing method mainly adopts time-domain prediction and time-domain coding methods, the frequency resolution is low, and the correlation between the auditory irrelevance and the frequency components of the audio signal cannot be removed well;

2. The signal processed by the psychoacoustic model is only transmitted to the receiving end as side information, not as part of the description encoding; so once the parameters in the psychoacoustic model are lost, the receiving end will not be able to understand the received audio. The signal is decoded correctly and thus less resistant to packet loss.

To sum up, since the second audio signal processing method only uses the psychoacoustic model with poor anti-packet loss performance as the basis of audio signal processing, the reliability of improving audio quality in the communication process is low, and it cannot effectively reduce the The problem of audio quality degradation in the communication process caused by packet loss and delay; moreover, audio quality degradation will significantly reduce user satisfaction.

Contents of the invention

In view of this, the main purpose of the present invention is to provide an audio signal processing method and system, so as to effectively improve audio quality during communication and improve user satisfaction.

Another object of the present invention is to provide an audio signal transceiving device, so as to effectively improve audio quality during communication and improve user satisfaction.

In order to achieve the above object, technical solution of the present invention is achieved in that way:

An audio signal processing method, comprising:

At the audio signal sending end, the residual signal and the masking threshold signal obtained by processing the audio signal are encoded into multiple residual signal descriptions and multi-channel masking threshold signal descriptions, and then the residual signal descriptions of each channel and the multi-channel masking threshold signal description are respectively In one path combination, the generated multiple paths all contain descriptions of residual signal description and masking threshold signal description;

At the audio signal receiving end, the remaining signal description and the masking threshold signal description contained in each of the received descriptions containing the description of the residual signal and the description of the masking threshold signal are branched, and then the branched Decode all the generated residual signal descriptions into one residual signal, and decode all the masking threshold signal descriptions generated after demultiplexing into one masking threshold signal, and perform parameter reconstruction and time-frequency analysis on the decoded residual signal and masking threshold signal Synthesis processing to generate a reconstructed audio signal.

The method of encoding the residual signal into a multiple residual signal description is:

Perform parity separation processing on the residual signal, and perform entropy coding processing on the multi-channel residual signal description generated by completing the processing;

The method of encoding the masking threshold signal into a multi-channel masking threshold signal description is:

Parity separation processing is performed on the masking threshold signal, and entropy coding processing is performed on the multi-channel masking threshold signal description generated by completing the processing.

The method of decoding all residual signal descriptions into one residual signal is:

Perform entropy decoding processing on all remaining signal descriptions, and perform parity synthesis processing on all remaining signal descriptions after decoding processing, and synthesize all remaining signal descriptions after parity synthesis processing into one residual signal;

The method of decoding all masking threshold signal descriptions into one masking threshold signal is:

Entropy decoding processing is performed on all masking threshold signal descriptions, parity synthesis processing is performed on all masking threshold signal descriptions after decoding processing, and all masking threshold signal descriptions after parity synthesis processing are synthesized into one masking threshold signal.

The method of encoding the residual signal into a multiple residual signal description is:

Perform signal pairing processing on the residual signal, and perform dual transformation on the multi-channel residual signal generated after the processing is completed to generate a corresponding multi-channel residual signal description, and then perform entropy coding on the generated multi-channel residual signal description deal with;

The method of encoding the masking threshold signal into a multi-channel masking threshold signal description is:

Perform signal pairing processing on the masking threshold signal, perform dual transformation on the multi-channel masking threshold signal generated after the processing, and generate a multi-channel masking threshold signal description corresponding to the number of channels, and then describe the generated multi-channel masking threshold signal Describes entropy encoding processing.

The method of decoding all residual signal descriptions into one residual signal is:

performing entropy decoding processing on all remaining signal descriptions, and performing dual inverse transform processing on all remaining signal descriptions after the decoding process is completed, and then synthesizing all remaining signal descriptions after completing the dual inverse transform processing into one residual signal;

The method of decoding all masking threshold signal descriptions into one masking threshold signal is:

Perform entropy decoding processing on all masking threshold signal descriptions, and perform dual inverse transform processing on all masking threshold signal descriptions after the decoding process, and then synthesize all masking threshold signal descriptions after completing the dual inverse transform processing into one masking threshold signal.

The multi-way description is a two-way description.

The method of encoding the residual signal into a multiple residual signal description is:

performing multi-description scalar quantization encoding processing on the residual signal, and performing entropy encoding processing on the multi-channel residual signal description generated after the processing;

The method of encoding the masking threshold signal into a multi-channel masking threshold signal description is:

Perform multi-description scalar quantization encoding processing on the masking threshold signal, and perform entropy encoding processing on the multi-channel masking threshold signal description generated after the processing.

The method of decoding all residual signal descriptions into one residual signal is:

Perform entropy decoding processing on all remaining signal descriptions, and perform multi-description scalar quantization decoding processing on all remaining signal descriptions after decoding processing, and decode all remaining signal descriptions after performing multi-description scalar quantization decoding processing into one residual signal;

The method of decoding all masking threshold signal descriptions into one masking threshold signal is:

Perform entropy decoding processing on all masking threshold signal descriptions, and perform multi-description scalar quantization decoding processing on all masking threshold signal descriptions after decoding processing, and decode all masking threshold signal descriptions after multi-description scalar quantization decoding processing into one masking threshold signal.

The multi-way description is more than two-way descriptions.

The residual signal is obtained after performing time-frequency analysis and residual signal analysis on the original audio signal;

The method of the time-frequency analysis is: performing processing including modified discrete cosine transform (MDCT) on the original audio signal to obtain time-frequency transform parameters;

The method for analyzing the residual signal is: removing the auditory irrelevant information or irrelevant degree in the time-frequency transformation parameters.

The masking threshold signal is obtained by analyzing the original audio signal with a psychoacoustic model.

Further, lossless encoding and audio packet processing are performed on the multi-channel residual signal description and the multi-channel masking threshold signal description generated by the sending end.

Before splitting at the receiving end, the audio packet unpacking and distortion-free decoding are further performed.

An audio signal processing system, the system includes a multi-description encoder located at the audio signal sending end, composed of a residual signal multi-description encoder and a masking threshold signal multi-description encoder all connected to a combiner; The multi-description decoder at the end is composed of a multi-description decoder for residual signals and a multi-description decoder for masked threshold signals, both of which are connected to the splitter;

Among them, the residual signal/concealment threshold signal multi-description encoder is used to encode the received residual signal into a multi-channel residual signal description, encode the received masking threshold signal into a multi-channel masking threshold signal description, and convert the encoded The generated descriptions are sent to the combiner;

A combiner, configured to respectively combine the received residual signal descriptions with one of the masking threshold signal descriptions to generate multiple descriptions that include both the residual signal description and the masking threshold signal description;

A splitter, configured to split the remaining signal descriptions and masking threshold signal descriptions contained in each of the received descriptions containing the remaining signal descriptions and masking threshold signal descriptions, and split the splitting All generated residual signal descriptions and masked threshold signal descriptions are sent to the residual signal/masked threshold signal multiple description decoder;

A residual signal/concealed threshold signal multi-description decoder, configured to decode all received residual signal descriptions into one residual signal, and decode all received masked threshold signal descriptions into one concealed threshold signal;

The multiple description decoder is further connected to a parameter reconstruction module connected to a time-frequency synthesis module;

Among them, the parameter reconstruction module is used to receive the residual signal and the masking threshold signal generated by the multi-description decoder, and perform parameter reconstruction processing on the received signal, and then send the time-frequency transformation parameters generated after the processing to Time-frequency synthesis module;

The time-frequency synthesis module is used to perform time-frequency synthesis processing on the received time-frequency transformation parameters to generate reconstructed audio signals.

The residual signal/masking threshold signal multi-descriptive encoder includes a connected parity separation module and an entropy encoder;

Wherein, the parity separation module is configured to perform parity separation processing on the residual signal, and send the multi-channel residual signal description generated by completing the processing to the entropy encoder, and perform parity separation processing on the masked threshold signal, and sending the description of the multi-channel masking threshold signal generated by completing the processing to the entropy encoder;

An entropy coder is used for performing entropy coding processing on the received multi-channel residual signal description and multi-channel masking threshold signal description.

The residual signal/masking threshold signal multi-description decoder includes a connected entropy decoder and a parity synthesis module;

Among them, the entropy decoder is used to perform entropy decoding processing on all remaining signal descriptions, and send all remaining signal descriptions after decoding processing to the parity synthesis module, perform entropy decoding processing on all masking threshold signal descriptions, and complete decoding All processed masking threshold signal descriptions are sent to the parity synthesis module;

The parity synthesis module is used to perform parity synthesis processing on all received residual signal descriptions, synthesize all residual signal descriptions after parity synthesis processing into one residual signal, and perform parity synthesis processing on all received masking threshold signal descriptions, The descriptions of all the masking threshold signals after parity combining processing are combined into one masking threshold signal.

The residual signal/masking threshold signal multi-description encoder includes a residual signal/masking threshold signal pairing module, a dual transformation module, and an entropy encoder connected in sequence;

Wherein, the residual signal/masking threshold signal pairing module is used to perform signal pairing processing on the residual signal, and send the multi-channel residual signal generated after the processing to the multi-description dual transformation module, and process the masking threshold signal Perform signal pairing processing, and send the multi-channel masking threshold signal generated by completing the processing to the multiple description dual transformation module;

The multi-description dual transformation module is used to perform dual transformation on the received multi-channel residual signals and generate corresponding multi-channel residual signal descriptions, and then send the generated multi-channel residual signal descriptions to the entropy encoder, and the received multi-channel residual signal descriptions Dual transformation is performed on the received multi-channel masking threshold signals respectively to generate corresponding multi-channel masking threshold signal descriptions, and then the generated multi-channel masking threshold signal descriptions are sent to the entropy encoder;

An entropy coder is used for performing entropy coding processing on the received multi-channel residual signal description and multi-channel masking threshold signal description.

The residual signal/masking threshold signal multi-description decoder includes an entropy decoder, a dual inverse transform decoder and a residual signal/masking threshold signal synthesis module connected in sequence;

Among them, the entropy decoder is used to perform entropy decoding processing on all remaining signal descriptions, and send all remaining signal descriptions generated after the decoding process to the multi-description dual inverse transform decoder, and perform entropy decoding on all masked threshold signal descriptions. Decoding processing, and sending all masked threshold signal descriptions after the decoding processing to the multiple description dual inverse transform decoder;

The multi-description dual inverse transform decoder is used to perform dual inverse transform processing on all remaining signal descriptions received, and send all remaining signal descriptions after the dual inverse transform processing to the residual signal/masking threshold signal synthesis module for receiving Perform dual inverse transform processing on all received masking threshold signal descriptions, and send all masking threshold signal descriptions after dual inverse transform processing to the remaining signal/masking threshold signal synthesis module;

The residual signal/masking threshold signal synthesis module is used to synthesize all received residual signal descriptions into one residual signal, and synthesize all received masking threshold signal descriptions into one masking threshold signal.

The residual signal/masking threshold signal multi-description encoder includes a connected multi-description scalar quantizer and an entropy encoder;

Wherein, the multi-description scalar quantizer is used to perform multi-description scalar quantization and encoding processing on the residual signal, and send the multi-channel residual signal description generated after the processing to the entropy encoder to perform multi-description on the masked threshold signal Scalar quantization encoding processing, and sending the multi-channel masking threshold signal description generated by the processing to the entropy encoder;

An entropy coder is used for performing entropy coding processing on the received multi-channel residual signal description and multi-channel masking threshold signal description.

The residual signal/masking threshold signal multi-description decoder includes a connected entropy decoder and a multi-description scalar quantization decoder;

Wherein, the entropy decoder is used to perform entropy decoding processing on all remaining signal descriptions, and send all remaining signal descriptions after decoding processing to the multi-description scalar quantization decoder, perform entropy decoding processing on all masked threshold signal descriptions, and Send all masked threshold signal descriptions after decoding processing to the multi-description scalar quantization decoder;

The multi-description scalar quantization decoder is used to perform multi-description scalar quantization and decoding processing on all received residual signal descriptions, decode all residual signal descriptions after multi-description scalar quantization and decoding processing into one residual signal, and decode all received The masking threshold signal description is subjected to multi-description scalar quantization decoding processing, and all masking threshold signal descriptions after the multi-description scalar quantization decoding processing are decoded into one masking threshold signal.

The multiple description scalar quantizer is a multiple description scalar quantizer capable of processing two or more descriptions, and the multiple description scalar quantization decoder is a multiple description scalar quantization decoder capable of processing two or more descriptions.

The multi-description encoder is further connected to the residual signal analysis module connected to the time-frequency analysis module, and also connected to the psychoacoustic module;

Wherein, the time-frequency analysis module is used for performing time-frequency analysis on the original audio signal, and sending the time-frequency transformation parameters generated after the analysis to the remaining signal analysis module;

A residual signal analysis module, configured to perform residual signal analysis on received time-frequency transformation parameters and masking threshold signals, and send the residual signal generated after the analysis to the multi-description encoder;

The psychoacoustic module is configured to perform psychoacoustic model analysis on the original audio signal, and send the masking threshold signal generated after analysis to the multiple description encoder and residual signal analysis module.

The multi-description encoder is further connected with the lossless encoding and audio packet processing module;

The lossless encoding and audio packet processing module is used to receive the description generated by the multi-description encoder, and perform lossless encoding and audio packet processing on the received description.

The multi-description decoder is further connected with the audio packet unpacking and distortion-free decoding module;

The audio packet unpacking and lossless decoding module is used for unpacking the audio packet and decoding without distortion for the received multi-channel description.

An audio signal sending device, which is composed of a multi-descriptive coder for residual signals and a multi-descriptive coder for masking threshold signals, both of which are connected to a combiner;

Residual signal/masking threshold signal multi-description encoder, used to encode the received residual signal into multiple residual signal descriptions, encode the received masking threshold signal into multiple masking threshold signal descriptions, and encode the generated The description of each channel is sent to the combiner;

The combiner is used to respectively combine the received residual signal descriptions with one of all masking threshold signal descriptions to generate multiple channels of descriptions that both include the residual signal descriptions and the masking threshold signal descriptions.

The residual signal/masking threshold signal multi-descriptive encoder includes a connected parity separation module and an entropy encoder;

Wherein, the parity separation module is configured to perform parity separation processing on the residual signal, and send the multi-channel residual signal description generated by completing the processing to the entropy encoder, and perform parity separation processing on the masked threshold signal, and sending the description of the multi-channel masking threshold signal generated by completing the processing to the entropy encoder;

An entropy coder is used for performing entropy coding processing on the received multi-channel residual signal description and multi-channel masking threshold signal description.

The residual signal/masking threshold signal multi-description encoder includes a residual signal/masking threshold signal pairing module, a multi-describing dual transformation module, and an entropy encoder that are sequentially connected;

Wherein, the residual signal/masking threshold signal pairing module is used to perform signal pairing processing on the residual signal, and send the multi-channel residual signal generated after the processing to the multi-description dual transformation module, and process the masking threshold signal Perform signal pairing processing, and send the multi-channel masking threshold signal generated by completing the processing to the multiple description dual transformation module;

The multi-description dual transformation module is used to perform dual transformation on the received multi-channel residual signals and generate corresponding multi-channel residual signal descriptions, and then send the generated multi-channel residual signal descriptions to the entropy encoder, and the received multi-channel residual signal descriptions Dual transformation is performed on the received multi-channel masking threshold signals respectively to generate corresponding multi-channel masking threshold signal descriptions, and then the generated multi-channel masking threshold signal descriptions are sent to the entropy encoder;

An entropy coder is used for performing entropy coding processing on the received multi-channel residual signal description and multi-channel masking threshold signal description.

The residual signal/masking threshold signal multi-description encoder includes a connected multi-description scalar quantizer and an entropy encoder;

Wherein, the multi-description scalar quantizer is used to perform multi-description scalar quantization and encoding processing on the residual signal, and send the multi-channel residual signal description generated after the processing to the entropy encoder to perform multi-description on the masked threshold signal Scalar quantization encoding processing, and sending the multi-channel masking threshold signal description generated by the processing to the entropy encoder;

An entropy coder is used for performing entropy coding processing on the received multi-channel residual signal description and multi-channel masking threshold signal description.

The multiple description scalar quantizer is a multiple description scalar quantizer capable of processing more than two descriptions.

The multi-description encoder is further connected to the residual signal analysis module connected to the time-frequency analysis module, and also connected to the psychoacoustic module;

Wherein, the time-frequency analysis module is used for performing time-frequency analysis on the original audio signal, and sending the time-frequency transformation parameters generated after the analysis to the remaining signal analysis module;

A residual signal analysis module, configured to perform residual signal analysis on received time-frequency transformation parameters and masking threshold signals, and send the residual signal generated after the analysis to the multi-description encoder;

The psychoacoustic module is configured to perform psychoacoustic model analysis on the original audio signal, and send the masking threshold signal generated after analysis to the multiple description encoder and residual signal analysis module.

The multi-description encoder is further connected with the lossless encoding and audio packet processing module;

The lossless encoding and audio package processing module is used to receive the multi-channel description generated by the multi-description encoder, and perform lossless encoding and audio package processing on the received multi-channel description.

An audio signal receiving device, which is composed of a multi-descriptive decoder for residual signals and a multi-descriptive decoder for masking threshold signals, both of which are connected to splitters;

Wherein, the splitter is used for splitting the remaining signal description and the masking threshold signal description included in each of the received descriptions containing the description of the remaining signal and the description of the masking threshold signal, and splitting the All remaining signal descriptions and masking threshold signal descriptions generated after are sent to the residual signal/masking threshold signal multi-description decoder;

A residual signal/concealed threshold signal multi-description decoder, configured to decode all received residual signal descriptions into one residual signal, and decode all received masked threshold signal descriptions into one concealed threshold signal;

The multiple description decoder is further connected to a parameter reconstruction module connected to a time-frequency synthesis module;

Among them, the parameter reconstruction module is used to receive the residual signal and the masking threshold signal generated by the multi-description decoder, and perform parameter reconstruction processing on the received signal, and then send the time-frequency transformation parameters generated after the processing to Time-frequency synthesis module;

The time-frequency synthesis module is used to perform time-frequency synthesis processing on the received time-frequency transformation parameters to generate reconstructed audio signals.

The residual signal/masking threshold signal multi-descriptive decoder includes connected entropy decoders and parity synthesis modules;

Among them, the entropy decoder is used to perform entropy decoding processing on all remaining signal descriptions, and send all remaining signal descriptions after decoding processing to the parity synthesis module, perform entropy decoding processing on all masking threshold signal descriptions, and complete decoding All processed masking threshold signal descriptions are sent to the parity synthesis module;

The parity synthesis module is used to perform parity synthesis processing on all received residual signal descriptions, synthesize all residual signal descriptions after parity synthesis processing into one residual signal, and perform parity synthesis processing on all received masking threshold signal descriptions, Respectfully, all masking threshold signals after parity combining processing are described and synthesized into one masking threshold signal.

Description of drawings

The residual signal/masking threshold signal multi-description decoder includes an entropy decoder, a multi-description dual inverse transform decoder and a residual signal/masking threshold signal synthesis module connected in sequence;

Among them, the entropy decoder is used to perform entropy decoding processing on all remaining signal descriptions, and send all remaining signal descriptions generated after the decoding process to the multiple description dual inverse transform decoder, and perform entropy decoding on all masking threshold signal descriptions processing, and send all masked threshold signal descriptions after decoding processing to the multiple description dual inverse transform decoder;

The multi-description dual inverse transform decoder is used to perform dual inverse transform processing on all remaining signal descriptions received, and send all remaining signal descriptions after the dual inverse transform processing to the residual signal/masking threshold signal synthesis module for receiving Perform dual inverse transform processing on all received masking threshold signal descriptions, and send all masking threshold signal descriptions after dual inverse transform processing to the remaining signal/masking threshold signal synthesis module;

The residual signal/masking threshold signal synthesis module is used to synthesize all received residual signal descriptions into one residual signal, and synthesize all received masking threshold signal descriptions into one masking threshold signal.

The residual signal/masking threshold signal multi-description decoder includes a connected entropy decoder and a multi-description scalar quantization decoder;

Wherein, the entropy decoder is used to perform entropy decoding processing on all remaining signal descriptions, and send all remaining signal descriptions after decoding processing to the multi-description scalar quantization decoder, perform entropy decoding processing on all masked threshold signal descriptions, and Send all masked threshold signal descriptions after decoding processing to the multi-description scalar quantization decoder;

The multi-description scalar quantization decoder is used to perform multi-description scalar quantization and decoding processing on all received residual signal descriptions, decode all residual signal descriptions after multi-description scalar quantization and decoding processing into one residual signal, and decode all received The masking threshold signal description is subjected to multi-description scalar quantization decoding processing, and all masking threshold signal descriptions after multi-description scalar quantization decoding processing are decoded into one masking threshold signal.

The multiple description scalar quantization decoder is a multiple description scalar quantization decoder capable of processing two or more descriptions.

The multi-description decoder is further connected with the audio packet unpacking and distortion-free decoding module;

The audio packet unpacking and distortion-free decoding module is used to unpack the audio packet and decode the received description without distortion.

Compared with the prior art, the audio signal processing method, system and audio signal transceiving device provided by the present invention, at the audio signal sending end, respectively encode the residual signal and the masking threshold signal obtained from the audio signal processing into the residual signal. More description and masking threshold signal multiple descriptions, and then respectively combine the remaining signal descriptions of each channel with one of the multi-channel masking threshold signal descriptions to generate a multi-channel description that includes both the residual signal and the masking threshold signal; at the audio signal receiving end, the The remaining signal descriptions and masking threshold signal descriptions contained in each description in all the received descriptions are split, and then all the remaining signal descriptions generated after splitting are decoded into one residual signal, and all the generated after splitting The masking threshold signal description is decoded into one masking threshold signal.

It can be seen that the present invention can effectively improve the audio quality in the communication process and improve user satisfaction.

Fig. 1 is the audio signal coder structure and schematic diagram of a preferred embodiment of the present invention;

Fig. 2 is the audio signal decoder structure and schematic diagram of a preferred embodiment of the present invention;

Fig. 3 is the multi-describing encoder structure and schematic diagram of the present invention;

Fig. 4 is the multi-descriptive decoder structure and schematic diagram of the present invention;

Fig. 5 is the structural and schematic diagram of the multi-descriptive decoder for the residual signal of the present invention;

Fig. 6 is the structure and schematic diagram of the masking threshold signal multi-describing decoder of the present invention;

Fig. 7 is the structural and schematic diagram of the residual signal multi-describing encoder of the present invention;

Fig. 8 is the multi-description decoder structure and schematic diagram that cooperates with the multi-description encoder in Fig. 7;

Fig. 9 is a structural and schematic diagram of a residual signal multi-describing encoder of the present invention;

Fig. 10 is the multi-description decoder structure and schematic diagram that cooperates with the multi-description encoder in Fig. 9;

Fig. 11 is a schematic diagram of audio signal processing in a preferred embodiment of the present invention;

Fig. 12 is a schematic diagram of a quantization interval in a preferred embodiment of the present invention;

Fig. 13 is a schematic diagram of a comparison of single-description and multiple-description quantization intervals in a preferred embodiment of the present invention;

Fig. 14 is a multi-description encoder structure and schematic diagram based on Fig. 11;

Fig. 15 is a multi-description decoder structure and schematic diagram based on Fig. 11;

Fig. 16 is a schematic diagram of audio signal processing in another preferred embodiment of the present invention;

Fig. 17 is a multi-description encoder structure and schematic diagram based on Fig. 16;

Fig. 18 is a multi-description decoder structure and schematic diagram based on Fig. 16;

Fig. 19 is a stereo diagram of encoding and decoding of a multi-description scalar quantizer according to a preferred embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below in conjunction with the drawings and specific embodiments.

In the audio signal processing method provided by the present invention, at the audio signal sending end, the residual signal and the masking threshold signal obtained by processing the audio signal are respectively coded into multiple descriptions of the residual signal and multiple descriptions of the masking threshold signal, and then the residual signals of each channel are respectively encoded into The description is combined with one of the multi-channel masking threshold signal descriptions to generate a multi-channel description that contains both the residual signal and the masking threshold signal; at the audio signal receiving end, the residual signal contained in each description in all received descriptions Describing and masking threshold signal descriptions are split, and then all remaining signal descriptions generated after splitting are decoded into one residual signal, and all masking threshold signal descriptions generated after splitting are decoded into one masking threshold signal.

The audio signal processing system provided by the present invention includes a multi-description encoder located at the audio signal sending end, composed of a residual signal multi-description encoder and a masking threshold signal multi-description encoder all connected to the combiner; The multi-description decoder at the receiving end is composed of a residual signal multi-description decoder and a masked threshold signal multi-description decoder all connected to the splitter;

Among them, the residual signal/masked threshold signal multi-description encoder is used to encode the received residual signal/masked threshold signal into a residual signal/masked threshold signal multi-description, and send the multi-description generated after encoding to the combiner The combiner is used to respectively combine the received residual signal descriptions with one of the masked threshold signal descriptions to generate multiple descriptions that contain the residual signal and the masked threshold signal; the splitter is used to combine The residual signal description and masking threshold signal description contained in each description in all received descriptions are divided into branches, and the residual signal/masking threshold signal multi-description generated after splitting is sent to the residual signal/masking threshold signal multi-description decoding The residual signal/masking threshold signal multi-description decoder is used to decode the received residual signal/masking threshold signal multi-describing into one path of residual signal/masking threshold signal.

The audio signal sending device provided by the present invention is composed of residual signal multi-description encoders and masking threshold signal multi-description encoders which are both connected to the combiner; wherein, the residual signal/masking threshold signal multi-description encoders are used to The received residual signal/masking threshold signal is coded into multiple descriptions of the residual signal/masking threshold signal, and the multiple description generated after encoding is sent to the combiner; the combiner is used to separately describe the received residual signals Combined with one path in all masking threshold signal descriptions to generate a multiplexed description that includes both the remaining signal and the masking threshold signal.

The audio signal receiving device provided by the present invention is composed of a multi-description decoder for residual signals and a multi-description decoder for masked threshold signals, both of which are connected to splitters; wherein, the splitter is used to divide each of the received descriptions into The residual signal description and masking threshold signal description contained in the description are divided into branches, and the residual signal/masking threshold signal multi-description generated after the splitting is sent to the residual signal/masking threshold signal multi-description decoder; the residual signal/masking threshold signal The multi-description decoder is used for multi-description decoding the received residual signal/masking threshold signal into one path of residual signal/masking threshold signal.

In principle, the present invention can be roughly divided into two levels. At the level of audio signal processing: it is necessary to analyze and synthesize the audio signal with multiple descriptions, such as: decomposing the audio signal into mutually irrelevant masking threshold signals and residual signals. ; At the level of quantization and coding: perform multiple description codec processing with multiple description and multiple decoders on the masked threshold signal and the residual signal respectively. Furthermore, when the channel packet loss is serious, the error concealment of the packet loss can also be carried out according to the historical records of different descriptions.

Referring to FIG. 1, FIG. 1 is a structural and schematic diagram of an audio signal encoder according to a preferred embodiment of the present invention. In FIG. 1 , the window-selected original audio signal is divided into two channels, one of which is input to the time-frequency analysis module 110 , and the other is input to the psychoacoustics module 120 .

The psychoacoustic module 120 performs psychoacoustic model analysis on the received audio signal to obtain a masking threshold signal related to the currently received audio frame, and sends the obtained masking threshold signal to the residual signal analysis module 130 and the multi-description encoder respectively. 140 and bit allocation module 150. The time-frequency analysis module 110 performs time-frequency analysis processing such as Modified Discrete Cosine Transform (MDCT) on the received audio signal, and sends time-frequency transformation parameters such as frequency domain MDCT coefficients obtained by the time-frequency analysis processing to the residual signal analysis module 130 ; The residual signal analysis module 130 uses the received masking threshold signal to remove the auditory irrelevant information or degree of incoherence in the received time-frequency transformation parameters, obtains the residual signal from which the auditory irrelevance has been removed and sends it to the multi-description encoder 140 .

The multi-description encoder 140 performs multi-description encoding on the received residual signal and the masking threshold signal that can represent the information of the current audio signal, to obtain two descriptions that can be processed separately or jointly: description 1 and description 2, and description 1 and description 2 to the lossless encoding and audio packet processing module 160. The bit allocation module 150 uses the received masking threshold signal as the control information of bit allocation, determines side information such as bit allocation mode, description identifier and quantizer identifier according to the masking threshold signal, and sends the determined side information to the lossless coding and Audio package processing module 160.

Lossless encoding and audio packet processing module 160 performs lossless encoding processes such as Huffman encoding, arithmetic encoding or run-length encoding to the received description 1 and description 2, to eliminate the redundancy of information sources and further reduce the bit rate; The side information is added to the description 1 and description 2 after encoding processing and bit-encapsulated, and then the encoded bit streams of the encapsulated description 1 and description 2 are sent to the channel.

Referring to FIG. 2, FIG. 2 is a structural and schematic diagram of an audio signal decoder in a preferred embodiment of the present invention. In FIG. 2 , an audio packet unpacking and distortion-free decoding module 210 , a multi-description decoder 220 , a parameter reconstruction module 230 and a time-frequency synthesis module 240 are connected in sequence. Among them, the audio packet unpacking and lossless decoding module 210 performs unpacking and decoding operations on the received description 1 and description 2, and sends the obtained description 1, description 2 and side information to the multi-description decoder 220; The decoder 220 performs multi-description decoding on the received description 1, description 2 and side information, and sends the decoded masking threshold signal and the remaining signal to the parameter reconstruction module 230; the parameter reconstruction module 230 utilizes the received masking threshold The time-frequency transformation parameters are reconstructed from the signal and the remaining signal, and the time-frequency transformation parameters obtained after the reconstruction are completed are sent to the time-frequency synthesis module 240, and the audio signal is reconstructed by the time-frequency synthesis module 240 according to the time-frequency transformation parameters.

In above-mentioned Fig. 1, Fig. 2, psychoacoustic model is set in the psychoacoustic module 120, and this model describes the perceptual characteristic of human ear to audio signal, and this description is mainly embodied in masking characteristic; From the perspective of audio compression and coding, The psychoacoustic model determines the energy of the maximum quantization noise that cannot be perceived by the human ear in the critical frequency band, or the noise masking threshold in the sense of auditory perception. Specifically, the psychoacoustic model can have different implementation methods, such as: adopting the model 1 adopted by the first and second layers of audio coding of MPEG-1 and MPEG-2 commonly used at present, or adopting the model 1 used as the third layer The model 2 used in MP3 audio coding, or the base curve (Floor) in Ogg Vorbis audio coding, can also use other types of psychoacoustic models such as the psychoacoustic model in AC3 audio coding.

The function of the time-frequency analysis module 110 is to transform or filter the time-domain audio signal, so as to remove the redundancy brought about by the correlation in the original audio signal. Transform-based time-frequency analysis can use time-frequency transformation methods such as MDCT, modulated lapped transform (MLT) or discrete wavelet transform (DWT), and transform-based time-frequency analysis obtains audio parameters in the transform domain or frequency domain; filter-based Time-frequency analysis can adopt the sub-band filtering algorithm similar to MPEG-1 and MPEG-2 audio coding, and perform time-frequency transformation processing such as MDCT, MLT or DWT in each sub-band, and obtain the transform domain or frequency Audio parameters for the domain. In addition, the function of the time-frequency synthesis module 240 in FIG. 2 is opposite to that of the time-frequency analysis module 110 , that is, the frequency-domain audio parameters are used for inverse transformation to obtain the reconstructed audio signal.

The main function of the remaining signal analysis module 130 is to remove the auditory irrelevance remaining in the frequency domain audio signal after the time-frequency analysis. If this processing is performed in the linear domain, the residual signal can be obtained by dividing the audio frequency domain parameters by the masking threshold signal; if it is performed in the logarithmic domain, the residual signal can be obtained by subtracting the masking threshold signal from the audio frequency domain parameters. In addition, the function of the parameter reconstruction module 230 in Fig. 2 is opposite to the function of the residual signal analysis module 130, that is: the residual signal obtained by multi-description decoding and the masking threshold signal are used to reconstruct the audio frequency domain parameters. If it is in the linear domain, the time-frequency transformation parameters can be obtained by multiplying the auditory residual signal and the masking threshold signal; if it is in the logarithmic domain, the time-frequency transformation parameters can be obtained by adding the residual signal and the masking threshold signal.

The function of the bit allocation module 150 is to control the quantization precision of the multi-description coding quantizer according to the received masking threshold signal, and at the same time perform dynamic bit allocation to the distortion-free coding and the formation of audio packets according to the available bits, and repeatedly adjust the quantization precision with an iterative method and bits are allocated until the number of available bits is exhausted, or the preset encoding quality has been reached. In practical applications, the bit allocation module 150 supports coding modes such as constant rate (CBR), variable rate (VBR) and average rate (ABR).

The lossless encoding and audio packet processing module 160 is used to perform lossless entropy encoding on description 1 and description 2 respectively, and then add side information to form two encoded description bit stream outputs. The two output description bit streams may be equal or not equal in importance, the number of bits required for the encoding of the above two descriptions may be the same or different, and the encoding rates of the two description bit streams Can be the same or different. In addition, the functions of the audio packet unpacking and lossless decoding module 210 are opposite to those of the lossless encoding and audio packet processing module 160, that is, two description bit streams are unpacked and lossless decoded to obtain two audio description information .

In practical applications, the multi-description encoder 140 shown in FIG. 1 may be shown in FIG. 3 , and the multi-description decoder 220 shown in FIG. 2 may be shown in FIG. 4 .

Referring to FIG. 3 , FIG. 3 is a structure and principle diagram of a multi-descriptive encoder according to the present invention. In FIG. 3 , the multi-description encoder 310 for the residual signal and the multi-description encoder 320 for the masked threshold signal are connected to a combiner 330 and a combiner 340 respectively. In practical application, the residual signal multi-description encoder 310 encodes the received residual signal, and sends the residual signal description 1 and residual signal description 2 formed by the encoding process to the combiner 330 and the combiner 340 respectively; The masking threshold signal multiple description encoder 320 encodes the received masking threshold signal, and sends the masking threshold signal description 1 and masking threshold signal description 2 formed by the encoding process to the combiner 330 and the combiner 340 respectively. The combiner 330 performs combination processing on the received residual signal description 1 and the masking threshold signal description 1, and sends the generated description 1 after the combination is completed; the combiner 340 combines the received residual signal description 2 and The description 2 of the masked threshold signal is combined, and the description 2 generated after the combination is completed is sent out.

There are many encoding algorithms that can be used by the multi-description encoder 310 for the residual signal and the multi-description encoder 320 for the masked threshold signal, such as: the currently commonly used multi-description scalar quantization algorithm (MDSQ), the multi-description transform coding algorithm (MDTC) or the multi-description Vector quantization (VQ) method. It is worth noting that: compared with the remaining signal, since the masking threshold signal may only contain a small amount of data, the multiple description encoding method corresponding to the masking threshold signal can also be a direct copy.

Of course, the multi-description encoder 310 for the residual signal and the multi-description encoder 320 for the masked threshold signal can also receive and process the side information, and then send the processed side information to the combiner 330 and the combiner 340, and the combiner 330, The combiner 340 combines the received side signals and other descriptions.

Referring to FIG. 4, FIG. 4 is a structural and schematic diagram of a multi-descriptive decoder in the present invention. In FIG. 4 , the splitter 410 and the splitter 420 are respectively connected to the residual signal multi-description decoder 430 and the masked threshold signal multi-description decoder 440 . In practical application, the demultiplexer 410 performs demultiplexing processing on the received description 1, and sends the residual signal descriptive 1 and the masking threshold signal descriptive 1 formed by the demultiplexing processing to the residual signal multi-description decoder 430, the masking threshold Signal multi-description decoder 440; splitter 420 splits the received description 2, and sends the remaining signal description 2 and masked threshold signal description 2 formed by the splitting process to the remaining signal multi-description decoder 430 , The masking threshold signal multi-descriptor decoder 440 . The residual signal multi-description decoder 430 decodes the received residual signal description 1 and residual signal description 2, and sends out the reconstructed residual signal generated after decoding; the masked threshold signal multi-description decoder 440 pairs The received masking threshold signal description 1 and masking threshold signal description 2 are decoded, and the reconstructed masking threshold signal generated after decoding is sent out.

Of course, the splitter 410 and the splitter 420 can also receive and process the side information, and then send the processed side information to the residual signal multiple description decoder 430 and the masking threshold signal multiple description decoder 440, and the residual signal multiple description The decoder 430 and the masked threshold signal multi-description decoder 440 decode the received side signals and other descriptions. Furthermore, in practical applications, only description 1 or description 2 may be sent to the splitter; in this case, the splitter that receives the description will normally perform subsequent processing such as splitting on the description , the completion of the branch description will also be followed by the normal decoding process.

In practical applications, the residual signal multi-description decoder 430 shown in FIG. 4 may be shown in FIG. 5 , and the masked threshold signal multi-description decoder 440 may be shown in FIG. 6 .

Referring to Fig. 5, Fig. 5 is a structure and schematic diagram of a residual signal multi-description decoder according to the present invention. The decoder shown in Fig. 5 adopts a three-decoder structure to perform multi-description decoding on received descriptions. Specifically, if only one description is received, the description is decoded by the residual signal multi-description decoder a 510 or the residual signal multi-description decoder c 530 as a side decoder receiving the description; if two A description, the multi-description decoder b 520 decodes the description with the remaining signal received as the central decoder.

Referring to Fig. 6, Fig. 6 is a structure and schematic diagram of a multi-description decoder for a masked threshold signal according to the present invention. The decoder shown in Fig. 6 adopts a three-decoder structure to perform multi-description decoding on received descriptions. Specifically, if only one description is received, the description is decoded with the masked threshold signal multi-description decoder a 610 or the masked threshold signal multi-description decoder c 630 that has received the description as a side decoder; When two descriptions are received, the description is decoded by the multi-description decoder b 620 that receives the description as the masked threshold signal of the central decoder.

The switch position of the output signal of the multi-description decoder shown in Fig. 5 and Fig. 6 can be automatically selected according to the situation of the received description.

In practical applications, the residual signal and masked threshold signal multi-description codec in Fig. 3 and Fig. 4 can be realized by multi-description parity separation algorithm, as shown in Fig. 7 and Fig. 8; it can also be realized by multi-description dual transformation algorithm , as shown in FIG. 9 and FIG. 10 ; it can also be realized by a multi-description scalar quantization algorithm, as shown in FIG. 11 to FIG. 18 .

In the following, the accompanying drawings are taken as an example to describe the above-mentioned different multi-description encoding and decoding algorithms respectively. It should be noted that, in the following description, the object of multi-description coding and decoding is mainly the residual signal; in practical applications, the same multi-description coding and decoding algorithm can also be applied to process masking threshold signals or other audio signal components.

For the case where the multi-description codec is realized by the multi-description parity separation algorithm, its operating principle is: the time-domain or frequency-domain audio parameters are separated according to their index values or natural order. Since the parity-separated two descriptions are completely uncorrelated; therefore the redundancy introduced between the two descriptions is zero and the overall encoding rate will not increase as a result.

Referring to FIG. 7, FIG. 7 is a structural and schematic diagram of a multi-descriptive encoder for residual signals according to the present invention. In Fig. 7, the parity separation module 710 performs parity separation processing on the remaining signals received, and sends the description 1 and description 2 generated by completing the processing to the entropy encoder 720 and the entropy encoder 730 respectively; The received description 1 is encoded, and the generated description 1 bit stream is sent after the encoding process is completed; the entropy encoder 730 is encoded for the received description 2, and the generated description 2 is generated after the encoding process is completed The bit stream is sent out.

Specifically, the residual signal is represented by R(k): R(k), k=1, 2, 3,...N

Wherein, N is the number of remaining signals (usually an even number), and is also half of the audio analysis window length.

The two multi-description algorithm signals of the remaining signal are represented by two descriptions M ₁ (k ₁ ) and M ₂ (k ₂ ) respectively:

Description 1: M ₁ (k ₁ ), k ₁ =1, 2, 3, ... N/2

Description 2: M ₂ (k ₂ ), k ₂ =1, 2, 3, ... N/2

Then the odd-even multi-description transformation algorithm and the results are as follows:

When k is an odd number, namely k=1, 3, 5, ... N-1, M ₁ (k ₁ )=R(k)

Among them, k ₁ =(k+1)/2;

When k is an even number, namely k=2, 4, 6, ... N, M ₂ (k ₂ )=R(k)

Among them, k ₂ =k/2.

After the above odd-even multi-description transformation, the entropy encoders 720 and 730 respectively encode the multi-description signals M ₁ (k ₁ ) and M ₂ (k ₂ ), and combine the encoded data with their respective masking threshold signals to form a description The bitstream is sent out.

Referring to FIG. 8 , FIG. 8 is a structural and schematic diagram of a multi-description decoder cooperating with the multi-description encoder in FIG. 7 . In Fig. 8, the entropy decoder 810 decodes the bit stream of the received description 1, and sends the generated description 1 to the parity synthesis module 830 and the parity synthesis module 840 after completing the decoding process; The received bit stream of description 2 is decoded, and the description 2 generated after the decoding process is completed is sent to the parity synthesis module 850 and the parity synthesis module 840.

The switch position of the output signal of the parity synthesis module shown in Figure 8 can be automatically selected according to the situation of receiving the description, such as: when only one description is received, the parity synthesis module that receives the description is used to decode the description , and connected to the output end of the parity synthesis module to output the remaining signal after the parity synthesis to complete the reconstruction; when both descriptions are received, the parity synthesis module 840 that receives the two descriptions The two descriptions are decoded and connected to the output terminal of the parity combination module 840 to output the reconstructed residual signal generated after the parity combination.

Specifically, if both descriptions are received, the two descriptions are decoded separately to obtain two multi-description signals: description 1 (M ₁ ) and description 2 (M ₂ ), and these two descriptions are synthesized into The residual signal, the specific synthesis algorithm is as follows:

When k is an odd number, that is, k=1, 3, 5, ... N-1, make R(k)=M ₁ (k ₁ ), where k ₁ =(k+1)/2;

When k is an even number, that is, k=2, 4, 6, ... N, make R(k)=M ₂ (k ₂ ), where k ₂ =k/2.

If only one description is received, this description is decoded to obtain a multi-description signal.

When this multiple description signal is description 1(M ₁ ):

When k is an odd number, namely k=1, 3, 5, ... N-1, make R(k)=M ₁ (k ₁ ), where k ₁ =(k+1)/2,

When k is an even number, that is, k=2, 4, 6, . . . N, set R(k)=0, where k ₂ =k/2.

When this description signal is description 2(M ₂ ):

When k is an odd number, that is, when k=1, 3, 5, ... N-1, make R(k)=0, wherein k ₁ =(k+1)/2,

When k is an even number, that is, k=2, 4, 6, ... N, make R(k)=M ₂ (k ₂ ), where k ₂ =k/2.

So far, the remaining signal has been successfully reconstructed.

For the case of multi-description encoding and decoding realized by the multi-description dual transformation algorithm, its operating principle is: through the transformation matrix T, two irrelevant variables A and B are transformed into two variables C and D with certain correlation. The magnitude of the correlation between variables C and D is determined by the transformation matrix T. Specifically, the transformation matrix is generally divided into an orthogonal matrix and a non-orthogonal matrix. The ranges of correlations introduced by the two transformation matrices are different, and the implementation methods of the corresponding dual transformation algorithms are also completely different.

In the following, only non-orthogonal matrices are taken as an example to describe the multiple description dual transformation algorithm.

The definition of each transformation parameter in the dual transformation algorithm is:

Input matrix: $\left[\begin{array}{c}A\\ B\end{array}\right]$

Dual transformation matrix: $T=\left[\begin{array}{cc}a& b\\ c& d\end{array}\right]$

Dual transformation output matrix: $\left[\begin{array}{c}C\\ \mathrm{D.}\end{array}\right]=T\left[\begin{array}{c}A\\ B\end{array}\right]$

Among them, a, b, c, d are the coefficients of the transformation matrix, which determine the redundancy introduced between the two descriptions, and ad-bc=1; A, B are input signals; C, D are output signals .

The specific algorithm of integer transformation is as follows: At the encoding end, the encoder performs dual positive transformation on signals A and B:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>\frac{\mathrm{<span\; class="notranslate">\; A</span>}}{\mathrm{<span\; class="notranslate">\; Q</span>}}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; ,</span>$ $\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>\frac{\mathrm{<span\; class="notranslate">\; B</span>}}{\mathrm{<span\; class="notranslate">\; Q</span>}}<span\; class="notranslate">\; ]</span>$

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; [</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; c</span>}}{\mathrm{<span\; class="notranslate">\; d</span>}}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; ]</span>$

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; wxya</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}$

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; [</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; b</span>}}{\mathrm{<span\; class="notranslate">\; d</span>}}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; ]</span>$

和 

分别为A和B量化后的整型变量， 

和 

分别为两个整型输出变量， 

和 

将被分别编码以形成两个描述比特流。Among them, Q is the quantization step size, W is an intermediate variable, and the symbol '[]' represents the rounding operation.

and

are the quantized integer variables of A and B respectively,

and

are two integer output variables,

and

will be encoded separately to form two description bitstreams.

At the decoding end, the specific decoding process has the following three situations depending on the received description:

和 

两个描述都被正确接收时，由收到这两个描述的解码器对信号 

和 进行下面对偶反变换：(1)

and

When both descriptions are received correctly, the signal is signaled by the decoder receiving both descriptions

and Perform the following dual inverse transformation:

$\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; +</span><span\; class="notranslate">\; [</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; b</span>}}{\mathrm{<span\; class="notranslate">\; d</span>}}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; ]</span>$

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; wxya</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}$

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; [</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; c</span>}}{\mathrm{<span\; class="notranslate">\; d</span>}}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; ]</span>$

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; ,</span>$ $\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}\mathrm{<span\; class="notranslate">\; Q</span>}$

分别为 

和 

逆量化后的值；‘[]’符号表示取整操作。Wherein, Q is the quantization step size; W is an intermediate variable; and

respectively

and

The value after inverse quantization; the '[]' symbol indicates a rounding operation.

被正确接收时，首先对丢失的信号 

进行预测：(2) Only one description

When correctly received, the missing signal is first

Make predictions:

$\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}\mathrm{<span\; class="notranslate">\; Q</span>}$

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="c"\; filenames="H061A3571320060801E000051.png,H061A3571320060801E000053.png"\; state="\{\{state\}\}">c</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="q"\; filenames="H061A3571320060801E000051.png,H061A3571320060801E000053.png"\; state="\{\{state\}\}">q</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; *</span>\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; \&phi;</span>}$

和 

进行反变换：again

and

Do the inverse transformation:

$\left[\begin{array}{c}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}\\ \mathrm{<span\; class="notranslate">\; B</span>}\end{array}\right]<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; T</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\left[\begin{array}{c}\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}\\ \stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}\end{array}\right]$

为 

经过解码器逆量化后的值； 

为解码器对 

预测恢复出来的值；矩阵T^-1为矩阵T的逆矩阵； 和 为解码器重构出来值；cosφ是变量C、D之间的相关系数。(3)只有一个描述 

被正确接收时，首先对丢失的信号 

进行预测，Among them, σ _c , σ _d and σ _q are the standard deviations of variables C, D and quantization error, respectively;

for

The value after inverse quantization by the decoder;

for the decoder pair

Predict the recovered value; matrix T ^-1 is the inverse matrix of matrix T; and is the value reconstructed by the decoder; cosφ is the correlation coefficient between variables C and D. (3) Only one description

When correctly received, the missing signal is first

make predictions,

$\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}\mathrm{<span\; class="notranslate">\; Q</span>}$

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; c</span>}}<span\; class="notranslate">\; *</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="c"\; filenames="H061A3571320060801E000051.png,H061A3571320060801E000053.png"\; state="\{\{state\}\}">c</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; <figure-callout\; id="2"\; label="q"\; filenames="H061A3571320060801E000051.png,H061A3571320060801E000053.png"\; state="\{\{state\}\}">q</figure-callout></span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; *</span>\stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; cos</span>}\mathrm{<span\; class="notranslate">\; \&phi;</span>}$

进行反变换：again and

Do the inverse transformation:

$\left[\begin{array}{c}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; A</span>}}\\ \stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; T</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\left[\begin{array}{c}\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; C</span>}}\\ \stackrel{<span\; class="notranslate">\; ~</span>}{\mathrm{<span\; class="notranslate">\; D.</span>}}\end{array}\right]$

预测恢复出来的值；矩阵T^-1为矩阵T的逆矩阵； 和 为解码器重构出来值；cosφ是变量C、D之间的相关系数。Among them, σ _c , σ _d and σ _q are the standard deviations of variables C, D and quantization error, respectively; for The value after dequantization by the decoder;

Predict the recovered value; matrix T ^-1 is the inverse matrix of matrix T; and is the value reconstructed by the decoder; cosφ is the correlation coefficient between variables C and D.

而 

是把量化后的A、B经过整型变换得到的。It should be noted that: C and D are obtained by multiplying the input signals A and B directly with the matrix T:

and

It is obtained by transforming quantized A and B through integer transformation.

Referring to Fig. 9, Fig. 9 is a structural and principle diagram of a multi-descriptive encoder for residual signals according to the present invention. In Fig. 9, the remaining signal pairing module 910 performs signal pairing processing on the received remaining signals, and sends the two-way remaining signals generated by completing the processing to the dual transformation module 920; The remaining signals are subjected to dual transformation processing respectively to generate two residual signals describing M ₁ and M ₂ , and then M ₁ and M ₂ are sent to the entropy encoder 930 and the entropy encoder ₉₄₀ respectively; Perform encoding processing, and send the bit stream of _M1 generated after the encoding process is completed; the entropy encoder 940 performs encoding processing on the received _M2 , and send the bit stream of _M2 generated after the encoding process is completed go out.

Specifically, the residual signal is represented by R(k): R(k), k=1, 2, 3,...N

Wherein, N is the number of remaining signals (N is generally an even number), and is also half of the audio analysis window length.

The two multi-description algorithm signals of the remaining signal are represented by two descriptions M ₁ (k ₁ ) and M ₂ (k ₂ ) respectively:

Description 1: M ₁ (k ₁ ), k ₁ =1, 2, 3, ... N/2

Description 2: M ₂ (k ₂ ), k ₂ =1, 2, 3, ... N/2

The principle shown in Figure 9 can be expressed as the following steps:

Step 1. Initialize the loop variable k=1 and other parameters of the dual transformation.

Step 2, the residual signal pairing module 910 performs pairing processing on the received residual signal, generates variables A and B and sends them to the dual transformation module 920; wherein, A=R(k), B=R(k+1) .

以及与描述2对应的 

Step 3, the dual transformation module 920 performs multiple description dual transformation on A and B, and obtains the corresponding to description 1

and corresponding to description 2

$m_1\left(k_1\right)=\stackrel{\&OverBar;}{C},$ where k ₁ =(k+1)/2,

$m_2\left(k_2\right)=\stackrel{\&OverBar;}{\mathrm{D.}},$ where k ₂ =(k+1)/2;

发送给熵编码器930，将 

发送给熵编码器940。Of course, the

Sent to the entropy encoder 930, the

Sent to entropy encoder 940.

Step 4. Let k=k+2, if k<N-1, go to step 2; otherwise, go to step 5.

Step 5. The entropy encoders 930 and 940 respectively encode the received multi-description signals M ₁ (k ₁ ) and M ₂ (k ₂ ), and form the described bits together with the encoded data and the respective masking threshold signals The stream is sent out.

It should be noted that in the dual transformation of multiple descriptions, the transformation matrix plays an important role, which controls the redundancy introduced between the two descriptions.

Referring to FIG. 10 , FIG. 10 is a structural and schematic diagram of a multi-description decoder cooperating with the multi-description encoder in FIG. 9 . In FIG. 10 , the entropy decoder 1001 decodes the received bit stream of description 1, and sends the generated description 1 after decoding to the dual inverse transform side decoder 1003 and the dual inverse transform central decoder 1004; The entropy decoder 1002 decodes the received bit stream of description 2, and sends the generated description 2 to the dual inverse transform side decoder 1005 and the dual inverse transform central decoder 1004 after the decoding process is completed. Furthermore, both the dual inverse transform side decoder and the dual inverse transform central decoder send the residual signal obtained by their own processing and description to the residual signal synthesis module; the residual signal synthesis module synthesizes the received residual signal to generate a synthesized remaining signal.

The switch position of the output signal of the residual signal synthesis module shown in Figure 10 can be selected automatically according to the situation of receiving the description, such as: when only one description is received, the dual inverse transform side decoder that receives the description is used to pair The description is decoded, and connected to the output end of the residual signal synthesis module connected to the dual inverse transform side decoder, so as to output the reconstructed residual signal generated after synthesis; when both descriptions are received, the Decode these two descriptions with the dual inverse transform center decoder 1004 that receives these two descriptions, and connect the output end of the residual signal synthesis module connected to the dual inverse transform side decoder 1004, to output the generated after synthesis The remaining signal of the completed reconstruction.

The specific decoding algorithm can be expressed as the following steps:

Step 1. Initialize the loop variable k=1 and other parameters of the dual transformation.

Step 2. Depending on the number of received descriptions, different decoding algorithms may be applied to decode the received descriptions. Specifically, the following processing is performed according to the received description:

(1) If the entropy decoders 1001 and 1002 have received two descriptions: M ₁ and M ₂ , then the entropy decoders receiving the above descriptions respectively determine the corresponding

$\stackrel{\&OverBar;}{C}=m_1\left(k_1\right),$ where k ₁ =(k+1)/2,

$\stackrel{\&OverBar;}{\mathrm{D.}}=m_2\left(k_2\right),$ where k ₂ =(k+1)/2.

和 

发送给对偶反变换中心解码器1004，由对偶反变换中心解码器1004根据多描述对偶变换算法求出 和 

并发送给剩余信号合成模块1007；之后，由剩余信号合成模块1007对收到的 和 进行合成处理，并将完成合成处理所生成的完成重构的剩余信号R(k)发送出去。Next, put

and

Sent to the dual inverse transform central decoder 1004, the dual inverse transform central decoder 1004 calculates and

And sent to the residual signal synthesis module 1007; Afterwards, the residual signal synthesis module 1007 performs synthesis processing on the received and , and sends out the reconstructed residual signal R(k) generated after the synthesis processing.

in, $R\left(k\right)=\hat{A},$ $R(k+1)=\hat{B}.$

( ₂ ) If only M ₁ is received, then the entropy decoder that receives M ₁ determines the corresponding

$\stackrel{\&OverBar;}{C}=m_1\left(k_1\right),$ where k ₁ =(k+1)/2,

发送给对偶反变换边解码器，对偶反变换边解码器根据多描述对偶变换算法预测出 并对 和 

进行对偶反变换求出 和 

再将其发送给剩余信号合成模块，由该剩余信号合成模块对收到的 和 进行合成处理，并将完成合成处理所生成的完成重构的剩余信号R(k)发送出去。The entropy decoder will

Sent to the dual inverse transform side decoder, the dual inverse transform side decoder predicts and to and

Perform dual inverse transformation to find and

It is then sent to the residual signal synthesis module, which performs synthesis processing on the received and , and sends out the reconstructed residual signal R(k) generated after the synthesis processing.

in, $R\left(k\right)=\hat{A},$ $R(k+1)=\hat{B}.$

( ₃ ) If only M ₂ is received, then the entropy decoder that receives M ₂ determines the corresponding

$\stackrel{\&OverBar;}{\mathrm{D.}}=m_2\left(k_2\right),$ where k ₂ =(k+1)/2.

发送给对偶反变换边解码器；对偶反变换边解码器根据多描述对偶变换算法预测出 

并对 

和 

进行对偶反变换求出 和 

再将其发送给剩余信号合成模块，由该剩余信号合成模块对收到的 和 进行合成处理，并将完成合成处理所生成的完成重构的剩余信号R(k)发送出去。The entropy decoder will

Sent to the dual inverse transform side decoder; the dual inverse transform side decoder predicts

and to

and

Perform dual inverse transformation to find and

It is then sent to the residual signal synthesis module, which performs synthesis processing on the received and , and sends out the reconstructed residual signal R(k) generated after the synthesis processing.

in, $R\left(k\right)=\hat{A},$ $R(k+1)=\hat{B}.$

Step 3, let k=k+2, if k<N-1, go to step 2; otherwise, go to step 4.

Step 4, end.

So far, the remaining signal has been successfully reconstructed.

For the multi-description codec implemented by the multi-description scalar quantization algorithm, its operating principle is: use a single-input multi-output scalar quantizer to quantize and encode the information source, and use the multiple output results obtained by quantization as the information Multiple descriptions of the source; a multi-input single-output scalar quantization decoder is used to encode the multi-description input, and the single output result obtained by decoding is used as the reconstructed signal of the source.

The principle and method of multi-description scalar quantization are described below by taking two descriptions as examples.

中心解码器1105用于对接收到的i和j进行解码，产生完成解码后的信号 

Referring to Fig. 11, Fig. 11 is a schematic diagram of audio signal processing in a preferred embodiment of the present invention. In Fig. 11, x is the original audio signal; l is the index obtained after quantization encoding; the matcher a(·) 1103 is used to match the index l into an index pair (i, j); the side decoders 1104, 1106 respectively Used to decode the received i and j to generate a decoded signal and

The central decoder 1105 is used to decode the received i and j to generate a decoded signal

In the whole process of multi-description scalar quantization, the most important thing is the realization of the matcher a(·)1103, that is, how to match l into (i, j). The following example illustrates:

After x undergoes encoding processing by the encoder 1102 , the range of the generated index l is as shown in FIG. 12 , which is: 1-10, that is, there are ten quantization intervals for quantizing x.

At this time, Table 1 can be used to realize the function of the matcher a(·)1103, that is, to match l into i and j:

Table 1 Multi-description scalar quantization table

The essence of the matching process realized by Table 1 is to replace the single-description quantizer (corresponding to the quantized index value l) with two multi-description quantizers (corresponding to the quantized index value i and j respectively), which The correspondence between the quantization intervals of the three quantizers is shown in FIG. 13 . In practical applications, the redundancy introduced between multiple descriptions can be controlled by adjusting the quantization precision of multiple descriptions. Usually, this adjustment can be realized by designing different tables. In general, the higher the quantization accuracy of multiple descriptions, the higher the redundancy introduced.

On the decoder side, the decoding process can take three specific forms:

(1) When receiving i and j, the central decoder 1105 can find a unique value of l according to Table 1, and can decode the signal l losslessly.

Specifically, the central decoder 1105 first finds the row and column corresponding to i and j in Table 1, and then finds the value at the intersection of the row and the column in Table 1, and quantizes the value as a multi-description scalar The previous value l.

(2) When only i is received, the decoder that receives i can estimate the value of l according to the data in Table 1. There are many methods of estimation, such as: take the average value of each row of data corresponding to i in Table 1 as the value of l; you can also use the maximum or minimum value of each row of data corresponding to i in Table 1 as The value of l.

(3) When only j is received, the specific decoding process is the same as when only i is received.

It should be noted that when designing a decoding algorithm, the optimal estimation method can be specifically determined according to the characteristics of the information source.

The encoding and decoding principle described in Figure 11 can be applied to Figures 14 and 15; in fact, the encoding and decoding process described in Figures 14 and 15 is only a specific application of the encoding and decoding principle in Figure 11.

Referring to FIG. 14 , FIG. 14 is a structure and schematic diagram of a multi-description encoder based on FIG. 11 . In Fig. 14, the dual-description scalar quantizer 1401 performs scalar quantization processing on the received residual signal, and sends the description 1 and description 2 generated after the processing to the entropy encoder 1402 and the entropy encoder 1403 respectively; the entropy encoder 1402 Encode the received description 1, and send the bit stream of description 1 generated after the encoding process is completed; the entropy encoder 1403 encodes the received description 2, and send the bit stream generated after the encoding process is completed. The bitstream describing 2 is sent.

Specifically, the residual signal is represented by R(k): R(k), k=1, 2, 3,...N

Wherein, N is the number of remaining signals, and is also half of the audio analysis window length.

The two multi-description algorithm signals of the remaining signal are represented by two descriptions M ₁ (k ₁ ) and M ₂ (k ₂ ) respectively:

M ₁ (k ₁ ), k ₁ =1, 2, 3, . . . N;

M ₂ (k ₂ ), k ₂ =1, 2, 3, . . . N.

In practical application, the principle shown in Figure 14 can be expressed as the following steps:

Step 1. Initialize the loop variable k: k=1.

Step 2. Using the remaining signal R(k) as an index value, search for a matching index pair M ₁ (k ₁ ), M ₂ (k ₂ ) according to Table 2; where k ₁ =k ₂ =k. The M ₁ (k ₁ ), M ₂ (k ₂ ) index pair are exactly the two description signals that need to be encoded, and this index pair is equivalent to the index pair (i, j) in FIG. 11 .

Table 2 Multi-description scalar quantization table

Step 3. Let k=k+1, if k<N, go to step 2; otherwise, go to step 4.

Step 4. The entropy encoder that receives M ₁ (k ₁ ) and M ₂ ( _{k 2} ) encodes M ₁ (k 1 ) and M ₂ (k ₂ ) respectively, and combines the encoded data with their respective The masking threshold signals together form the described bit stream and send it out.

In the process of multi-description scalar quantization, the design of Table 2 plays a key role. The smaller the difference between the data on the diagonal in Table 2, the higher the described quantization accuracy and the higher the encoding rate with the same sound quality.

The decoding process corresponding to the encoding process described in FIG. 14 is shown in FIG. 15 . Referring to FIG. 15 , FIG. 15 is a structural and schematic diagram of a multi-description decoder based on FIG. 11 . In Fig. 15, the entropy decoder 1501 decodes the received bit stream of description 1, and sends the description 1 generated after the decoding processing to the multi-description scalar quantization side decoder 1503 and the multi-description scalar quantization central decoder 1504 ; the entropy decoder 1502 decodes the received bit stream of description 2, and sends the generated description 2 to the multi-description scalar quantization side decoder 1505 and the multi-description scalar quantization central decoder 1504 .

The switch positions of the multi-description scalar quantization side decoder and the multi-description scalar quantization center decoder output signal shown in Figure 15 can be automatically selected according to the situation of the received description, such as: when only one description is received, the received The multi-description scalar quantization side decoder to the description performs scalar quantization decoding on the description, and connects the output end of the multi-description scalar quantization side decoder to output the reconstructed residual signal generated after decoding; when two When the descriptions are all received, the multi-description scalar quantization central decoder 1504 that receives the two descriptions is used to perform dual-description scalar quantization decoding on the two descriptions, and the output terminal of the multi-description scalar quantization central decoder 1504 is connected to output The reconstructed residual signal generated after decoding.

Specifically, the principle shown in Figure 15 can be expressed as the following steps:

Step 1. Set k=1.

Step 2. Process respectively according to the situation of receiving the description bit stream:

If two descriptions are received: M ₁ (k ₁ ) and M ₂ (k ₂ ), the value of the remaining signal R(k) can be found uniquely according to Table 2;

If only description 1: M ₁ (k ₁ ) is received, find the row corresponding to description 1 according to Table 2, and use the smallest absolute value in this row as the value of the remaining signal R(k);

If only description 2: M ₂ (k ₂ ) is received, find the column corresponding to description 2 according to Table 2, and use the minimum absolute value in this column as the value of the remaining signal R(k);

where k ₁ =k ₂ =k.

Step 3. Let k=k+1, if k<N, go to step 2; otherwise, go to step 4.

Step 4, end.

So far, the remaining signal has been successfully reconstructed.

The above is the principle and method of scalar quantization for the two descriptions; in practical applications, when the channel packet loss rate is greater than 25% or even higher, if the scalar quantization method of the two descriptions is still used, then when the two descriptions are intermittent Even when consecutive simultaneous losses occur, the audio continuity necessary for audio communication will be severely affected. In this case, if the number of descriptions can be increased, problems such as sound quality degradation and audio discontinuity caused by higher packet loss rates can be effectively avoided; it can be seen that the algorithm framework containing three or more descriptions is in It is often necessary in practical applications.

To this end, the following three descriptions are taken as examples to illustrate the principles and methods of multi-description scalar quantization; as for other types and algorithms with more descriptions, this can be used as a reference for design. Moreover, in these algorithm descriptions, the object of multi-description coding and decoding is mainly the residual signal; of course, in practical applications, the object of multi-description coding and decoding can also be the masking threshold signal or other audio signal components.

或 

双描述边解码器用于对接收到的h和i、h和j、i和j这样的两个描述进行解码，产生完成解码后的信号 

中心解码器用于对接收到的全部三个描述信号h、i和j进行解码，产生完成解码后的信号 

Referring to Fig. 16, Fig. 16 is a schematic diagram of audio signal processing in another preferred embodiment of the present invention. In Fig. 16, x is the original audio signal, and l is the index obtained after quantization and encoding; the matcher a(·) 1603 is used to match the index l into an index set (h, i, j); It is used to decode the received h, i or j, and generate the decoded signal

or

The dual description side decoder is used to decode the received two descriptions of h and i, h and j, i and j, and generate the decoded signal

The central decoder is used to decode all three received description signals h, i, and j to generate a decoded signal

The multi-description scalar quantization process shown in FIG. 16 is the same in principle as the multi-description scalar quantization process shown in FIG. 11 , and will not be repeated here. In fact, the principle of operation is the same whether it is a double-description, triple-description or more-description encoding and decoding process:

For the encoding process, according to the number of descriptions to be generated, the residual signal and the masking threshold signal are respectively encoded into the remaining signal description and the masking threshold signal description with the same number of descriptions, and then each residual signal description is combined with one of the The masked-threshold signal descriptions are combined; multiple descriptions are finally generated, and each description contains the residual signal and the masked-threshold signal.

For the decoding process, the remaining signal descriptions and masking threshold signal descriptions in the received descriptions are split, and then all the remaining signal descriptions generated after splitting are decoded into one residual signal, and the generated after splitting All masking threshold signal descriptions of are decoded into one masking threshold signal.

The encoding and decoding principle described in Figure 16 can be applied to Figures 17 and 18; in fact, the encoding and decoding process described in Figures 17 and 18 is only a specific application of the encoding and decoding principle in Figure 16.

Referring to FIG. 17 , FIG. 17 is a structure and principle diagram of a multi-description encoder based on FIG. 16 . In Fig. 17, the three-description scalar quantizer 1701 performs three-description scalar quantization processing on the received residual signal, and sends the description 1, description 2, and description 3 generated after the processing to the entropy encoder 1702 and the entropy encoder 1703 respectively . Entropy encoder 1704; entropy encoders 1702, 1703, and 1704 respectively encode the received description, and send out the description bit stream generated after the encoding process is completed.

Specifically, the remaining signal is represented by R(k): R(k), k=1, 2, 3,...N;

Wherein, N is the number of remaining signals, and is also half of the audio analysis window length.

The three multi-description algorithm signals of the remaining signal are respectively represented by two descriptions M ₁ (k ₁ ), M ₂ (k ₂ ) and M ₃ (k ₃ ):

M ₁ (k ₁ ), k ₁ =1, 2, 3, . . . N;

M ₂ (k ₂ ), k ₂ =1, 2, 3, ... N;

M ₃ (k ₃ ), k ₃ =1, 2, 3, . . . N.

In practical application, the principle shown in Figure 17 can be expressed as the following steps:

Step 1. Initialize the loop variable k: k=1.

Step 2. Use the remaining signal R(k) as an index value, and search for matching three-dimensional coordinate points M ₁ (k ₁ ), M ₂ (k ₂ ) and M ₃ (k ₃ ); where k ₁ =k ₂ =k ₃ = k. The three-dimensional coordinate points M ₁ (k ₁ ), M ₂ (k ₂ ) and M ₃ (k ₃ ) are just the three description signals that need to be encoded; and the three-dimensional coordinate points M ₁ (k ₁ ), M ₂ (k ₂ ) and M ₃ (k ₃ ) correspond to the three description signals (h, i, j) in FIG. 16 .

Step 3. Let k=k+1, if k<N, go to step 2; otherwise, go to step 4.

Step 4. Entropy encoders that receive M ₁ (k ₁ ), M ₂ (k ₂ ), and M ₃ (k ₃ ) respectively encode M ₁ (k ₁ ), M ₂ (k ₂ ), M ₃ (k 3 ₃ ) Perform coding processing, and send the coded data together with their respective masking threshold signals to form a described bit stream.

The decoding process corresponding to the encoding process described in FIG. 17 is shown in FIG. 18 . Referring to FIG. 18 , FIG. 18 is a structure and principle diagram of a multi-description decoder based on FIG. 16 . In Fig. 18, the entropy decoder 1801 decodes the received bit stream of description 1, and sends the generated description 1 to the multi-description scalar quantization side decoders 1804, 1806, 1809 and multi-description scalar Quantization center decoder 1807; entropy decoder 1802 decodes the received bitstream of description 2, and sends description 2 generated after decoding processing to multi-description scalar quantization side decoders 1805, 1806, 1808 and multiple Descriptor scalar quantization central decoder 1807; entropy decoder 1803 decodes the received bit stream of description 3, and sends description 3 generated after decoding processing to multi-description scalar quantization side decoders 1808, 1809, 1810 and multi-description scalar quantization center decoder 1807.

The switch positions of the multi-description scalar quantization side decoder and the multi-description scalar quantization center decoder output signal shown in Figure 18 can be automatically selected according to the situation of the received description, such as: when only one or two descriptions are received, Decode the description with the multi-description scalar quantization side decoder receiving the description, and connect the output end of the multi-description scalar quantization side decoder to output the reconstructed residual signal generated after decoding; when When all three descriptions are received, the multi-description scalar quantization central decoder 1807 that receives these three descriptions is used to decode the three descriptions, and the output port of the multi-description scalar quantization central decoder 1807 is connected to output the decoded The remaining signal generated after the completion of the reconstruction.

Specifically, the principle shown in Figure 18 can be expressed as the following steps:

Step 1. Set k=1.

Step 2. Process respectively according to the situation of receiving the description bit stream:

If three descriptions are received: M ₁ (k ₁ ), M ₂ (k ₂ ) and M ₃ (k ₃ ), the value of the remaining signal R(k) can be found uniquely according to the received descriptions;

If only two descriptions in M ₁ (k ₁ ), M ₂ (k ₂ ) and M ₃ (k ₃ ) are received, find the corresponding axis according to the received description, and put the axis with the smallest absolute value value as the value of the residual signal R(k);

If only one description is received: M ₁ (k ₁ ), M ₂ (k ₂ ) or M ₃ (k ₃ ), find the corresponding plane according to the received description, and take the value with the smallest absolute value on the plane as the value of the residual signal R(k);

where k ₁ =k ₂ =k ₃ =k.

Step 3. Let k=k+1, if k<N, go to step 2; otherwise, go to step 4.

Step 4, end.

So far, the remaining signal has been successfully reconstructed.

It should be noted that in the process of three-description scalar quantization, the design shown in Figure 19 plays a key role in finding the value of the remaining signal according to the description. Referring to Fig. 19, Fig. 19 is a three-dimensional diagram of encoding and decoding of a multi-description scalar quantizer according to a preferred embodiment of the present invention. In Fig. 19, three orthogonal coordinate axes respectively represent three descriptions: M ₁ (k ₁ ), M ₂ (k ₂ ), and M ₃ (k ₃ ). The design principle of Figure 19 is basically the same as that of Table 2, the only difference is that the diagonal line in Table 2 becomes a straight line at an angle of 60 degrees to the three coordinate axes, and the distance between the data on this straight line The smaller the difference, the less data is distributed around it, the less the number of remaining signal values corresponding to each described value, the higher the quantization accuracy of the description, and the higher the encoding rate of the same sound quality.

In practical applications, all the multi-description encoding and decoding algorithms mentioned above can guarantee: in the ideal case of no packet loss, the audio signal can be decoded normally by using the two descriptions received, and can be In the scope, one received description is used to estimate another lost description, and then the correlation between descriptions is used to recover and reconstruct the audio signal. However, when the packet loss rate continues to increase, two descriptions may be lost at the same time, and it is difficult for a multi-description decoder to perform audio decoding and audio reconstruction well. Therefore, in order to improve the sound quality when the packet loss is serious, the multi-description coding and decoding algorithm can be further improved and perfected; therefore, several packet loss concealment processing algorithms based on multi-description coding are proposed below:

1. In the aforementioned odd-even separation multi-description encoding algorithm, since the odd and even descriptions are completely independent two descriptions, there is no correlation between them; therefore, when one of the descriptions is lost, the received Another description estimates the missing description. To this end, the following packet loss concealment processing algorithm can be used:

Replace the description of the current frame loss with the description of the normal reception of the previous frame; or, multiply the description of the normal reception of the previous frame by an attenuation factor (can be set to: between 0.5 and 0.9), and replace the current frame with the multiplied value The description of frame loss; or, replace the description of current frame loss with the description of normal reception of the current frame and the linear interpolation of 0.

2. When two descriptions are lost at the same time due to severe packet loss, the descriptions received normally in the previous frame are multiplied by an attenuation factor (can be set between 0.5 and 0.9), and the multiplied values are used to replace The description of the current frame loss, and use it to estimate the audio parameters or audio signals of the current frame.

3. When a certain description is continuously lost due to severe packet loss, the frame-by-frame decrement algorithm is used, and the description of the last frame received normally is multiplied by an attenuation factor by frame (can be set to: between 0.5 and 0.9) , replace the description of the current frame loss with the multiplied value respectively.

In the above figures, if there is an entropy encoder/decoder, in fact, only one entropy encoder/decoder can be used to encode/decode the description, instead of two or more entropy encoders/decoders as shown in the figure and, the entropy encoder/decoder can also be replaced by another type of encoder/decoder; moreover, before processing the residual signal, it can be further rounded, quantized and/or encoded.

It can be seen from the above description that the audio signal processing method, system and audio signal transceiving device provided by the present invention can effectively improve the audio quality in the communication process and improve user satisfaction.

Claims (38)

1. An audio signal processing method characterized by:

at an audio signal sending end, coding a residual signal and a masking threshold signal obtained by processing the audio signal into a plurality of paths of residual signal descriptions and a plurality of paths of masking threshold signal descriptions respectively, and combining the residual signal descriptions and one path of the masking threshold signal descriptions respectively to generate a plurality of paths of descriptions which all comprise the residual signal descriptions and the masking threshold signal descriptions;

at an audio signal receiving end, branching all the received residual signal description and masking threshold signal description contained in each description in the descriptions which all contain residual signal description and masking threshold signal description, decoding all the residual signal descriptions generated after branching into a path of residual signal, decoding all the masking threshold signal descriptions generated after branching into a path of masking threshold signal, and performing parameter reconstruction and time-frequency synthesis processing on the residual signal and the masking threshold signal obtained after decoding to generate a reconstructed audio signal.

2. A method as claimed in claim 1, characterized in that the residual signal is encoded into the multi-residual signal description by:

carrying out odd-even separation processing on the residual signals, and carrying out entropy coding processing on the multi-channel residual signal description generated by the processing;

the method for encoding the masking threshold signal into a multi-path masking threshold signal description comprises the following steps:

and carrying out odd-even separation processing on the masking threshold signal, and carrying out entropy coding processing on the multi-path masking threshold signal description generated by the processing.

3. A method as claimed in claim 2, characterized in that the decoding of all residual signal descriptions into one residual signal is performed by:

entropy decoding all the residual signal descriptions, performing parity synthesis processing on all the residual signal descriptions after decoding processing, and synthesizing all the residual signal descriptions after parity synthesis processing into one residual signal;

the method for decoding all the masking threshold signal descriptions into a path of masking threshold signals comprises the following steps:

and performing entropy decoding processing on all the masking threshold signal descriptions, performing parity synthesis processing on all the masking threshold signal descriptions after the decoding processing is completed, and synthesizing all the masking threshold signal descriptions after the parity synthesis processing into a masking threshold signal.

4. A method as claimed in claim 1, characterized in that the residual signal is encoded into the multi-residual signal description by:

carrying out signal pair division processing on the residual signals, carrying out dual transformation on the multi-channel residual signals generated after the processing is finished, generating multi-channel residual signal descriptions with corresponding channels, and carrying out entropy coding processing on the generated multi-channel residual signal descriptions;

the method for encoding the masking threshold signal into a multi-path masking threshold signal description comprises the following steps:

and carrying out signal pair division processing on the masking threshold signals, carrying out dual transformation on the multi-path masking threshold signals generated after the processing is finished, generating multi-path masking threshold signal descriptions of corresponding paths, and carrying out entropy coding processing on the generated multi-path masking threshold signal descriptions.

5. The method of claim 4, wherein decoding all residual signal descriptions into one residual signal comprises:

entropy decoding all the residual signal descriptions, dual inverse transformation processing all the residual signal descriptions after decoding processing, and synthesizing all the residual signal descriptions after the dual inverse transformation processing into a path of residual signal;

the method for decoding all the masking threshold signal descriptions into a path of masking threshold signals comprises the following steps:

and performing entropy decoding processing on all the masking threshold signal descriptions, performing dual inverse transformation processing on all the masking threshold signal descriptions after the decoding processing is completed, and synthesizing all the masking threshold signal descriptions after the dual inverse transformation processing into a masking threshold signal.

6. The method of any of claims 2 to 5, wherein the multi-path description is a two-path description.

7. A method as claimed in claim 1, characterized in that the residual signal is encoded into the multi-residual signal description by:

carrying out multi-description scalar quantity quantization coding processing on the residual signals, and carrying out entropy coding processing on the multi-path residual signal description generated after the processing is finished;

the method for encoding the masking threshold signal into a multi-path masking threshold signal description comprises the following steps:

and carrying out multi-description scalar quantity quantization coding processing on the masking threshold signal, and carrying out entropy coding processing on the multi-path masking threshold signal description generated after the processing is finished.

8. The method of claim 7 wherein decoding all residual signal descriptions into a residual signal comprises:

entropy decoding all residual signal descriptions, performing multi-description scalar quantization decoding on all residual signal descriptions after decoding processing, and decoding all residual signal descriptions after the multi-description scalar quantization decoding processing into one path of residual signal;

the method for decoding all the masking threshold signal descriptions into a path of masking threshold signals comprises the following steps:

and performing entropy decoding processing on all the masking threshold signal descriptions, performing multi-description scalar quantity quantization decoding processing on all the masking threshold signal descriptions after the decoding processing is completed, and decoding all the masking threshold signal descriptions after the multi-description scalar quantity quantization decoding processing into one path of masking threshold signals.

9. The method of claim 7 or 8, wherein the multiple descriptions are two or more descriptions.

10. The method of claim 1, wherein the residual signal is obtained by performing time-frequency analysis on an original audio signal and residual signal analysis;

11. the method of claim 10, wherein:

the time-frequency analysis method comprises the following steps: processing the original audio signal including Modified Discrete Cosine Transform (MDCT) to obtain time-frequency transform parameters;

the method for analyzing the residual signal comprises the following steps: and removing auditory incoherent information or incoherent degree in the time-frequency transformation parameters.

12. The method of claim 1, wherein the masking threshold signal is obtained by performing a psychoacoustic model analysis on an original audio signal.

13. The method of claim 1, further comprising performing distortion-free encoding and audio packet processing on the plurality of residual signal descriptions and the plurality of masking threshold signal descriptions generated at the transmitting end.

14. The method of claim 13, wherein audio packet de-packetization and distortion-free decoding are further performed before splitting at the receiving end.

15. An audio signal processing system is characterized by comprising a multiple description encoder positioned at an audio signal transmitting end, a multi-description encoder for residual signals and a multi-description encoder for masking threshold signals, wherein the multiple description encoder for residual signals and the multi-description encoder for masking threshold signals are connected with a combiner; the multi-description decoder is positioned at the audio signal receiving end and consists of a residual signal multi-description decoder and a masking threshold signal multi-description decoder which are connected with the splitter;

the device comprises a residual signal/masking threshold signal multi-description encoder, a combiner and a residual signal/masking threshold signal multi-description encoder, wherein the residual signal/masking threshold signal multi-description encoder is used for encoding a received residual signal into a plurality of paths of residual signal descriptions, encoding a received masking threshold signal into a plurality of paths of masking threshold signal descriptions and sending each path of descriptions generated after encoding to the combiner;

the combiner is used for respectively combining the received residual signal descriptions with one of all the masking threshold signal descriptions to generate a plurality of paths of descriptions which all contain residual signal descriptions and masking threshold signal descriptions;

a splitter, configured to split the residual signal description and the masking threshold signal description contained in each of the descriptions that all contain the residual signal description and the masking threshold signal description, and send all the residual signal descriptions and the masking threshold signal descriptions generated after splitting to a residual signal/masking threshold signal multiple description decoder;

a residual signal/masking threshold signal multi-description decoder, for decoding all residual signal descriptions received into a path of residual signal, and decoding all masking threshold signal descriptions received into a path of masking threshold signal;

the multi-description decoder is further connected with a parameter reconstruction module connected with a time-frequency synthesis module;

the parameter reconstruction module is used for receiving the residual signal and the masking threshold signal generated by the multi-description decoder, performing parameter reconstruction processing on the received signal, and sending the time-frequency transformation parameter generated after the processing to the time-frequency synthesis module;

and the time-frequency synthesis module is used for carrying out time-frequency synthesis processing on the received time-frequency transformation parameters to generate a reconstructed audio signal.

16. The system of claim 15, wherein said residual signal/masking threshold signal multiple description coder comprises a connected parity separation module, an entropy coder;

the parity separation module is used for performing parity separation processing on the residual signals, sending the multi-channel residual signal description generated by the processing to the entropy coder, performing parity separation processing on the masking threshold signal, and sending the multi-channel masking threshold signal description generated by the processing to the entropy coder;

and the entropy coder is used for performing entropy coding processing on the received multi-path residual signal description and the multi-path masking threshold signal description.

17. The system of claim 16, wherein the residual signal/masking threshold signal multiple description decoder comprises a connected entropy decoder, parity synthesis module;

the entropy decoder is used for performing entropy decoding processing on all residual signal descriptions, sending all residual signal descriptions subjected to decoding processing to the parity synthesis module, performing entropy decoding processing on all masking threshold signal descriptions, and sending all masking threshold signal descriptions subjected to decoding processing to the parity synthesis module;

the parity synthesis module is used for performing parity synthesis processing on all the received residual signal descriptions, synthesizing all the residual signal descriptions subjected to the parity synthesis processing into a path of residual signal, performing parity synthesis processing on all the received masking threshold signal descriptions, and synthesizing all the masking threshold signal descriptions subjected to the parity synthesis processing into a path of masking threshold signal.

18. The system of claim 15, wherein said residual signal/masking threshold signal multiple description coder comprises a residual signal/masking threshold signal pair-dividing module, a dual transform module and an entropy coder connected in sequence;

the device comprises a residual signal/masking threshold signal pair division module, a multi-description dual transformation module and a multi-description dual transformation module, wherein the residual signal/masking threshold signal pair division module is used for carrying out signal pair division processing on the residual signal, sending the multi-path residual signal generated after the processing is finished to the multi-description dual transformation module, carrying out signal pair division processing on the masking threshold signal and sending the multi-path masking threshold signal generated after the processing is finished to the multi-description dual transformation module;

the multi-description dual transformation module is used for respectively carrying out dual transformation on the received multi-channel residual signals to generate multi-channel residual signal descriptions with corresponding paths, then sending the generated multi-channel residual signal descriptions to the entropy coder, respectively carrying out dual transformation on the received multi-channel masking threshold signals to generate multi-channel masking threshold signal descriptions with corresponding paths, and then sending the generated multi-channel masking threshold signal descriptions to the entropy coder;

and the entropy coder is used for performing entropy coding processing on the received multi-path residual signal description and the multi-path masking threshold signal description.

19. The system of claim 18, wherein the residual signal/masking threshold signal multiple description decoder comprises an entropy decoder, a dual inverse transform decoder, and a residual signal/masking threshold signal synthesis module connected in series;

the entropy decoder is used for performing entropy decoding processing on all residual signal descriptions, sending all residual signal descriptions generated after the decoding processing to the multi-description dual inverse transform decoder, performing entropy decoding processing on all masking threshold signal descriptions, and sending all the masking threshold signal descriptions after the decoding processing to the multi-description dual inverse transform decoder;

the multi-description dual inverse transformation decoder is used for carrying out dual inverse transformation processing on all received residual signal descriptions, sending all residual signal descriptions subjected to the dual inverse transformation processing to the residual signal/masking threshold signal synthesis module, carrying out dual inverse transformation processing on all received masking threshold signal descriptions, and sending all masking threshold signal descriptions subjected to the dual inverse transformation processing to the residual signal/masking threshold signal synthesis module;

and the residual signal/masking threshold signal synthesis module is used for describing and synthesizing all the received residual signals into a residual signal and describing and synthesizing all the received masking threshold signals into a masking threshold signal.

20. The system of claim 15, wherein said residual signal/masking threshold signal multiple description encoder comprises a concatenated multiple description scalar quantizer, entropy encoder;

the multi-description scalar quantizer is used for performing multi-description scalar quantization coding processing on the residual signals, sending the multi-path residual signal description generated after the processing to the entropy coder, performing multi-description scalar quantization coding processing on the masking threshold signals, and sending the multi-path masking threshold signal description generated after the processing to the entropy coder;

and the entropy coder is used for performing entropy coding processing on the received multi-path residual signal description and the multi-path masking threshold signal description.

21. The system of claim 20, wherein the residual signal/masking threshold signal multiple description decoder comprises a concatenated entropy decoder, multiple description scalar quantization decoder;

the entropy decoder is used for performing entropy decoding processing on all residual signal descriptions, sending all residual signal descriptions subjected to decoding processing to the multi-description scalar quantization decoder, performing entropy decoding processing on all masking threshold signal descriptions, and sending all masking threshold signal descriptions subjected to decoding processing to the multi-description scalar quantization decoder;

and the multi-description scalar quantization decoder is used for performing multi-description scalar quantization decoding processing on all the received residual signal descriptions, decoding all the residual signal descriptions subjected to the multi-description scalar quantization decoding processing into one path of residual signals, performing multi-description scalar quantization decoding processing on all the received masking threshold signal descriptions, and decoding all the masking threshold signal descriptions subjected to the multi-description scalar quantization decoding processing into one path of masking threshold signals.

22. The system of claim 20, wherein:

the multiple description scalar quantizer is a multiple description scalar quantizer capable of processing two or more descriptions, and the multiple description scalar quantization decoder is a multiple description scalar quantization decoder capable of processing two or more descriptions.

23. The system of claim 15, wherein the multiple description coder is further coupled to a residual signal analysis module coupled to a time-frequency analysis module, and to a psychoacoustic module;

the time-frequency analysis module is used for performing time-frequency analysis on the original audio signal and sending time-frequency transformation parameters generated after the analysis is finished to the residual signal analysis module;

a residual signal analysis module, configured to perform residual signal analysis on the received time-frequency transform parameter and masking threshold signal, and send a residual signal generated after the analysis to the multi-description encoder;

and the psychoacoustic module is used for performing psychoacoustic model analysis on the original audio signal and sending the masking threshold signal generated after the analysis to the multi-description encoder and the residual signal analysis module.

24. The system of any of claims 15 to 23, wherein the multiple description coder is further coupled to an undistorted coding and audio packet processing module;

and the undistorted coding and audio packet processing module is used for receiving the description generated by the multi-description coder and carrying out undistorted coding and audio packet processing on the received description.

25. The system of claim 24, wherein the multiple description decoder is further coupled to an audio packet unpacking and distortion-free decoding module;

and the audio packet unpacking and distortion-free decoding module is used for carrying out audio packet unpacking and distortion-free decoding processing on the received multi-channel description.

26. An audio signal transmitting device is characterized in that the device consists of a residual signal multi-description encoder and a masking threshold signal multi-description encoder which are connected with a combiner;

a residual signal/masking threshold signal multi-description encoder for encoding the received residual signal into a plurality of residual signal descriptions, encoding the received masking threshold signal into a plurality of masking threshold signal descriptions, and sending each path of description generated after encoding to a combiner;

and the combiner is used for respectively combining the received residual signal descriptions with one of all the masking threshold signal descriptions to generate a plurality of paths of descriptions which all contain residual signal descriptions and masking threshold signal descriptions.

27. The apparatus of claim 26, wherein said residual signal/masking threshold signal multiple description coder comprises a connected parity separation module, an entropy coder;

the parity separation module is used for performing parity separation processing on the residual signals, sending the multi-channel residual signal description generated by the processing to the entropy coder, performing parity separation processing on the masking threshold signal, and sending the multi-channel masking threshold signal description generated by the processing to the entropy coder;

and the entropy coder is used for performing entropy coding processing on the received multi-path residual signal description and the multi-path masking threshold signal description.

28. The apparatus of claim 26, wherein said residue signal/masking threshold signal multiple description coder comprises a residue signal/masking threshold signal pair division module, a multiple description dual transform module and an entropy coder connected in sequence;

the device comprises a residual signal/masking threshold signal pair division module, a multi-description dual transformation module and a multi-description dual transformation module, wherein the residual signal/masking threshold signal pair division module is used for carrying out signal pair division processing on the residual signal, sending the multi-path residual signal generated after the processing is finished to the multi-description dual transformation module, carrying out signal pair division processing on the masking threshold signal and sending the multi-path masking threshold signal generated after the processing is finished to the multi-description dual transformation module;

the multi-description dual transformation module is used for respectively carrying out dual transformation on the received multi-channel residual signals to generate multi-channel residual signal descriptions with corresponding paths, then sending the generated multi-channel residual signal descriptions to the entropy coder, respectively carrying out dual transformation on the received multi-channel masking threshold signals to generate multi-channel masking threshold signal descriptions with corresponding paths, and then sending the generated multi-channel masking threshold signal descriptions to the entropy coder;

and the entropy coder is used for performing entropy coding processing on the received multi-path residual signal description and the multi-path masking threshold signal description.

29. The apparatus of claim 26, wherein said residual signal/masking threshold signal multiple description encoder comprises a concatenated multiple description scalar quantizer, entropy encoder;

the multi-description scalar quantizer is used for performing multi-description scalar quantization coding processing on the residual signals, sending the multi-path residual signal description generated after the processing to the entropy coder, performing multi-description scalar quantization coding processing on the masking threshold signals, and sending the multi-path masking threshold signal description generated after the processing to the entropy coder;

and the entropy coder is used for performing entropy coding processing on the received multi-path residual signal description and the multi-path masking threshold signal description.

30. The apparatus of claim 29, wherein the multiple description scalar quantizer is a multiple description scalar quantizer capable of processing more than two descriptions.

31. The apparatus of claim 26, wherein said multiple description coder is further coupled to a residual signal analysis module coupled to a time-frequency analysis module, and to a psychoacoustic module;

the time-frequency analysis module is used for performing time-frequency analysis on the original audio signal and sending time-frequency transformation parameters generated after the analysis is finished to the residual signal analysis module;

a residual signal analysis module, configured to perform residual signal analysis on the received time-frequency transform parameter and masking threshold signal, and send a residual signal generated after the analysis to the multi-description encoder;

and the psychoacoustic module is used for performing psychoacoustic model analysis on the original audio signal and sending the masking threshold signal generated after the analysis to the multi-description encoder and the residual signal analysis module.

32. The apparatus of any of claims 26 to 31, wherein the multiple description coder is further coupled to an undistorted coding and audio packet processing module;

and the undistorted coding and audio packet processing module is used for receiving the multi-path description generated by the multi-description coder and carrying out undistorted coding and audio packet processing on the received multi-path description.

33. An audio signal receiving apparatus, characterized in that the apparatus is composed of a residual signal multiple description decoder and a masking threshold signal multiple description decoder both connected to a splitter;

the device comprises a splitter, a residual signal/masking threshold signal multi-description decoder and a residual signal/masking threshold signal multi-description decoder, wherein the splitter is used for splitting residual signal descriptions and masking threshold signal descriptions contained in all descriptions which contain residual signal descriptions and masking threshold signal descriptions and sending all residual signal descriptions and masking threshold signal descriptions generated after splitting to the residual signal/masking threshold signal multi-description decoder;

a residual signal/masking threshold signal multi-description decoder, for decoding all residual signal descriptions received into a path of residual signal, and decoding all masking threshold signal descriptions received into a path of masking threshold signal;

the multi-description decoder is further connected with a parameter reconstruction module connected with a time-frequency synthesis module;

the parameter reconstruction module is used for receiving the residual signal and the masking threshold signal generated by the multi-description decoder, performing parameter reconstruction processing on the received signal, and sending the time-frequency transformation parameter generated after the processing to the time-frequency synthesis module;

and the time-frequency synthesis module is used for carrying out time-frequency synthesis processing on the received time-frequency transformation parameters to generate a reconstructed audio signal.

34. The apparatus of claim 33, wherein the residual signal/masking threshold signal multiple description decoder comprises an entropy decoder, a parity synthesis module, connected;

the entropy decoder is used for performing entropy decoding processing on all residual signal descriptions, sending all residual signal descriptions subjected to decoding processing to the parity synthesis module, performing entropy decoding processing on all masking threshold signal descriptions, and sending all masking threshold signal descriptions subjected to decoding processing to the parity synthesis module;

and the parity synthesis module is used for performing parity synthesis processing on all the received residual signal descriptions, synthesizing all the residual signal descriptions subjected to the parity synthesis processing into a path of residual signal, performing parity synthesis processing on all the received masking threshold signal descriptions, and synthesizing all the masking threshold signal descriptions subjected to the parity synthesis processing into a path of masking threshold signal.

35. The apparatus of claim 33, wherein the residual signal/masking threshold signal multiple description decoder comprises an entropy decoder, a multiple description dual inverse transform decoder, and a residual signal/masking threshold signal synthesis module connected in sequence;

the entropy decoder is used for performing entropy decoding processing on all residual signal descriptions, sending all residual signal descriptions generated after the decoding processing to the multi-description dual inverse transform decoder, performing entropy decoding processing on all masking threshold signal descriptions, and sending all the masking threshold signal descriptions after the decoding processing to the multi-description dual inverse transform decoder;

the multi-description dual inverse transformation decoder is used for carrying out dual inverse transformation processing on all received residual signal descriptions, sending all residual signal descriptions subjected to the dual inverse transformation processing to the residual signal/masking threshold signal synthesis module, carrying out dual inverse transformation processing on all received masking threshold signal descriptions, and sending all masking threshold signal descriptions subjected to the dual inverse transformation processing to the residual signal/masking threshold signal synthesis module;

and the residual signal/masking threshold signal synthesis module is used for describing and synthesizing all the received residual signals into a residual signal and describing and synthesizing all the received masking threshold signals into a masking threshold signal.

36. The apparatus of claim 33, wherein the residual signal/masking threshold signal multiple description decoder comprises a concatenated entropy decoder, multiple description scalar quantization decoder;

the entropy decoder is used for performing entropy decoding processing on all residual signal descriptions, sending all residual signal descriptions subjected to decoding processing to the multi-description scalar quantization decoder, performing entropy decoding processing on all masking threshold signal descriptions, and sending all masking threshold signal descriptions subjected to decoding processing to the multi-description scalar quantization decoder;

and the multi-description scalar quantization decoder is used for performing multi-description scalar quantization decoding processing on all the received residual signal descriptions, decoding all the residual signal descriptions subjected to the multi-description scalar quantization decoding processing into one path of residual signals, performing multi-description scalar quantization decoding processing on all the received masking threshold signal descriptions, and decoding all the masking threshold signal descriptions subjected to the multi-description scalar quantization decoding processing into one path of masking threshold signals.

37. The apparatus of claim 36, wherein the multiple description scalar quantization decoder is a multiple description scalar quantization decoder capable of processing more than two descriptions.

38. The apparatus of claim 33, wherein the multiple description decoder is further coupled to an audio packet unpacking and distortion-free decoding module;

and the audio packet unpacking and distortion-free decoding module is used for carrying out audio packet unpacking and distortion-free decoding processing on the received description.

Images