CN101751925B - Tone decoding method and device - Google Patents

Abstract

The invention discloses a speech decoding method and device. The method includes generating the high frequency spectrum of the current frame by using the low frequency spectrum of the current frame which is a narrow frequency band; judging whether the narrowband frame needs to be corrected according to certain judgment conditions, and needs to be corrected , correct the high frequency spectrum of the narrowband frame; convert the current frame processed according to the judgment result from the frequency domain signal to the time domain signal; use the attenuation factor of the current frame After attenuation, the time domain signal of the current frame is output. Correspondingly, the present invention also discloses a voice decoding device, which includes an expansion unit, a first modification processing unit, a conversion unit and an attenuation unit. The embodiments of the present invention use a voice decoding method and device to well solve the problem of hearing discomfort caused by voice signals with different bandwidths, and improve the auditory quality of the voice signals.

Description

A voice decoding method and device

technical field

The invention relates to communication technology, in particular to a voice decoding method and device.

Background technique

For the current multi-rate voice and audio encoding and decoding algorithm, during the network transmission process of the coded code stream, due to the different network status, the network will truncate the code stream transmitted from the coder to the network with different lengths. At the decoding end, the decoder will decode voice and audio signals with different bandwidths or bit rates according to the truncated code stream, so that the received voice and audio signals will switch between different bandwidths.

Due to the relationship between the filter state parameters in the decoder and the front and rear voice and audio signals, when the bandwidth is switched, the filter state parameters will be discontinuous. If some appropriate processing is not done in time, there will be Some phenomena of sudden changes in energy will cause the quality of hearing to deteriorate; moreover, when different frequency bandwidths appear in continuous signals, it will also cause uncomfortable hearing effects.

There is a Bandwidth Extension (BWE) technology in the prior art, which extends the narrow-band signal to the wide-band signal in the following way: it copies the low-band spectrum of the current frame to the high-band, and then corrects it through some parameters The low-band spectrum of the current frame, so that the energy of the low-band spectrum is more in line with the energy of the high-band spectrum. At present, in the frequency band extension technology, generally some parameter information is obtained from the low frequency information of the current frame or a small amount of high frequency band parameters of the current frame are transmitted from the encoding end to the decoding end to restore the signal of the high frequency band part of the current frame . However, when the wideband speech signal is switched to a narrowband speech signal, the current frame of the narrowband speech signal does not have any information in the high frequency part, and the existing frequency band extension technology cannot well solve the above problems caused by different bandwidths. Problems with voice signals.

Contents of the invention

Embodiments of the present invention provide a voice decoding method and device, which can improve the auditory quality of voice signals and avoid the problem of hearing discomfort caused by voice signals with different bandwidths when decoding voice signals.

For this reason, the embodiment of the present invention provides a kind of speech decoding method, comprising:

generating the high frequency spectrum of the current frame by using the low frequency spectrum of the current frame which is a narrow frequency band;

According to the low frequency band information of the current frame and the frequency band information of the previous frame relative to the current frame, it is judged whether the current frame needs to be corrected. The frequency band spectrum is corrected, and if the judgment result is no, the current frame is not corrected;

converting the current frame processed according to the judgment result from a frequency domain signal to a time domain signal;

After attenuating the high-frequency time-domain signal of the time-domain signal of the current frame by using the attenuation factor of the current frame, the time-domain signal of the current frame is output.

Correspondingly, the embodiment of the present invention also provides a speech decoding device, including:

an extension unit, configured to generate the high frequency spectrum of the current frame by using the low frequency frequency spectrum of the current frame which is a narrow frequency band;

The first correction processing unit is used to judge whether the current frame needs to be corrected according to the low frequency band information of the current frame and the frequency band information of the previous frame relative to the current frame, and when the judgment result is yes, use the correction Correcting the high-frequency spectrum of the current frame by a factor, and not performing correction on the current frame when the judgment result is negative;

A transformation unit, configured to transform the current frame processed according to the judgment result from a frequency-domain signal to a time-domain signal;

The attenuation unit is configured to use the attenuation factor of the current frame to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame, and then output the time-domain signal of the current frame.

The embodiment of the present invention utilizes the low-band spectrum of the narrow-band frame to generate the high-band spectrum of the narrow-band frame, and judges whether the narrow-band frame needs to be corrected according to certain judgment conditions. The band spectrum enables the narrow-band frame in the voice signal to be expanded into an effective wide-band frame, which solves the problem of auditory discomfort caused by voice signals of different bandwidths and improves the auditory quality of the voice signal.

Description of drawings

Fig. 1 is a schematic flow chart of a speech decoding method according to Embodiment 1 of the present invention;

Fig. 2 is a schematic flow chart of a speech decoding method according to Embodiment 2 of the present invention;

Fig. 3 is a schematic flow chart of determining the attenuation factor of the current frame in the speech decoding method of the embodiment of the present invention;

4 is a schematic flow chart of a speech decoding method according to Embodiment 3 of the present invention;

FIG. 5 is a schematic structural diagram of a speech decoding device according to Embodiment 4 of the present invention;

Fig. 6 is a schematic structural diagram of a first correction processing unit in the speech decoding device of Fig. 5;

FIG. 7 is a schematic structural diagram of a speech decoding device according to Embodiment 5 of the present invention.

Detailed ways

The speech decoding method and device proposed by the embodiments of the present invention can improve the auditory quality of the speech signal, and avoid the problem of hearing discomfort caused by the speech signals of different bandwidths when the speech signal is decoded.

The technical solutions of the embodiments of the present invention will be described in detail below in the form of specific embodiments in conjunction with the accompanying drawings.

FIG. 1 is a flowchart of a speech decoding method according to Embodiment 1 of the present invention. As shown in Figure 1, the method includes:

Step S101, generating a high frequency spectrum of the current frame according to the low frequency frequency spectrum of the current frame which is a narrow frequency band.

Step S102 , judging whether the current frame needs to be corrected according to the low frequency band information of the current frame and the frequency band information of a previous frame relative to the current frame. If the judgment result is yes, execute step S103 and then execute step S104. If the judgment result is no, execute step S104.

In this embodiment, the previous frame mode bit and the current frame mode bit are set, and an initial value is set for the previous frame mode bit. Specifically, determine the value of the mode bit of the current frame according to the value of the mode bit of the previous frame, the low frequency band information of the current frame, and the frequency band information of the previous frame relative to the current frame; Whether the frame needs to be corrected.

Step S103 , correcting the high frequency spectrum of the current frame by using the correction factor.

For example, the high-band spectrum of the current frame is corrected by multiplying the high-band spectrum of the current frame by a correction factor.

In this embodiment, the value of the correction factor is specifically the average energy value of the low-band spectrum of the current frame; or the ratio of the generated current frame's high-band spectrum energy to the low-band spectrum energy multiplied by a decimal, so that The value of the correction factor is less than 1, so that the high frequency spectrum energy value of the current frame is less than a certain multiple of its low frequency spectrum energy value, optionally, the high frequency spectrum energy value of the current frame is less than its low frequency spectrum energy value 0.8 times the value.

For example, the way to calculate the high-band spectral energy value of a frame signal is: the mean value of the sum of the squares of the high-frequency spectral coefficients of the frame is the high-frequency spectral energy value of the frame;

Similarly, the way to calculate the low-band spectral energy value of a frame signal is: the mean value of the sum of the squares of the low-band spectral coefficients of the frame is the low-band spectral energy value of the frame;

The value obtained by comparing the high frequency spectral energy value of a frame signal with the low frequency spectral energy value of the frame is the ratio of the high frequency spectral energy to the low frequency spectral energy of the frame.

Step S104, converting the current frame from a frequency domain signal to a time domain signal.

Step S105 , using the attenuation factor of the current frame to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame.

In this embodiment, the attenuation factor of the current frame is determined according to the frame type information of the current frame, the frame type information of the previous frame relative to the current frame and its attenuation factor; for each frame after the current frame, the The attenuation factor of a frame is determined according to the frame type information of the frame, the frame type information of a previous frame relative to the frame and its attenuation factor. During implementation, the determined attenuation factor of the current frame is multiplied by the high-frequency time-domain signal of the time-domain signal of the current frame. Wherein, the frame type information is a wideband frame or a narrowband frame. The specific implementation mode is shown in Fig. 3 for details.

Step S106, outputting the time domain signal of the current frame. Here, the time-domain signal of the current frame includes its high-frequency time-domain signal and low-frequency time-domain signal.

This embodiment avoids the discontinuity of the filter state caused by the switching between speech signals of different bandwidths, resulting in an energy shock The phenomenon makes the decoded continuous speech signal more comfortable in hearing, and improves the subjective sound quality and auditory quality.

FIG. 2 is a schematic flowchart of a speech decoding method according to Embodiment 2 of the present invention. As shown in Figure 2, the received continuous speech signal is decoded, the method includes:

Step S201, judging whether the current frame is a wideband frame or a narrowband frame. If the current frame is a broadband frame, execute step S211. If the current frame is a narrowband frame, execute step S202.

Step S202, multiply the normalized low-band spectrum of the current frame by the high-band frequency domain envelope of the previous frame relative to the current frame and copy it to the high frequency, generate the high-band spectrum of the current frame, and convert the current frame Extend from narrowband frame to wideband frame.

Step S203 , judging whether the original frame relative to the previous frame of the current frame is a wideband frame or a narrowband frame. If the original frame of the previous frame is a broadband frame, step S204 is executed and then step S206 is executed. If the original frame of the previous frame is a narrowband frame, step S205 is executed and then step S207 is executed.

Step S204, according to the frequency band information of the previous frame, calculate the energy value of the low frequency band spectrum of the previous frame, the energy value of the high frequency band spectrum, and the energy ratio of the high frequency band spectrum and the low frequency band spectrum of the previous frame, and according to the current The low frequency band information of the frame is used to calculate the low frequency band spectrum energy value of the current frame, and according to the generated high frequency band information of the current frame, the high frequency band spectrum energy value of the current frame is calculated, and the high frequency band spectrum energy value of the current frame is calculated. The energy ratio of the low-band spectrum.

Wherein, since the high-band spectrum of the current frame is generated by using the high-band frequency envelope of the previous frame relative to the frame and copying the low-band spectrum of the current frame, the above-mentioned high Frequency band information, calculating the energy value of the high frequency band spectrum of the current frame refers to calculating the energy value of the current frame based on the energy value of the low frequency band spectrum of the current frame and the high frequency domain envelope value of the previous frame relative to the frame High frequency band spectral energy value.

In this embodiment, the energy of the low-band spectrum of the previous frame is denoted as E _{p1_L} , the energy of the high-band spectrum is denoted as E _{p1_H} , and the energy ratio of the high-band spectrum and low-band spectrum of the previous frame is denoted as E _gain1 ; record the energy of the low-band spectrum of the current frame as E _{c2_L} , the energy of the high-band spectrum as E _{c2_H} , and the energy ratio of the high-band spectrum and the low-band spectrum of the current frame as E _gain2 .

In this embodiment, the method of calculating the low frequency band spectral energy value, the high frequency band spectral energy value of a frame signal, and the ratio of the high frequency band spectral energy to the low frequency band spectral energy of a frame signal is the same as the steps in the first embodiment The implementation manner described in S103 is the same, so details are not repeated here.

Step S205 : Calculate the value of the low-band spectrum energy E _{p1_L} of the previous frame according to the low-band information of the previous frame, and calculate the value of the low-band spectrum energy E _{c2_L} of the current frame according to the low-band information of the current frame.

Step S206, according to the initial value of the mode bit of the previous frame, and the value of the low-band spectrum energy E _{p1_L} of the previous frame calculated in step S204, and the energy ratio E of the high-band spectrum and the low-band spectrum of the previous frame The value of _gain1 , the value of the low-band spectrum energy E _{c2_L} of the current frame, and the value of the energy ratio E _gain2 of the high-band spectrum and the low-band spectrum of the current frame determine the value of the mode bit of the current frame.

In this embodiment, the previous frame mode bit is recorded as prev_mode with an initial value of 0, and the current frame mode bit is recorded as mode. The steps are specifically:

If the current frame satisfies the following three conditions at the same time, set mode to 1. The three conditions are:

1. prev_mode=0;

2. The energy ratio E _gain1 of the high frequency spectrum and the low frequency spectrum of the previous frame is smaller than the energy ratio E _gain2 of the high frequency spectrum and the low frequency spectrum of the current frame;

3. The low-band spectrum energy E _{p1_L} of the previous frame is greater than the low-band spectrum energy E _{c2_L} of the current frame by a certain multiple α. In this embodiment, α is generally above 1.6.

Step S207, according to the value of the previous frame mode bit, and the value of the low-band spectral energy E _{p1_L} of the previous frame calculated in step S205, and the value of the low-band spectral energy E _{c2_L} of the current frame, determine the current frame mode bit value.

In this embodiment, the mode bit of the previous frame is recorded as prev_mode, and the mode bit of the current frame is recorded as mode. The steps are specifically:

If the frame meets the following two conditions at the same time, the mode position is set to 0. The two conditions are:

1. prev_mode=1;

2. The low-frequency spectral energy E _{c2_L} of the current frame is greater than the low-frequency spectral energy E _{p1_L} of the previous frame by a certain multiple δ. In this embodiment, δ is generally above 1.6.

Step S208, judging whether the value of the mode bit of the current frame is 1. If the value is 1, it is determined that the current frame needs to be corrected, step S210 is executed and then step S211 is executed. If the value is 0, it is determined that the current frame does not need to be corrected, and step S211 is directly executed.

Step S209, assigning the value of the mode bit of the current frame to the mode bit of the previous frame to update the value of the mode bit of the previous frame in step S207.

Step S210, using a correction factor to correct the high frequency spectrum of the current frame, so that its energy is closer to the real high frequency spectrum energy.

For example, the high-band spectrum of the current frame is corrected by multiplying the high-band spectrum of the current frame by a correction factor.

In this embodiment, the value of the correction factor is specifically the average energy value of the low-band spectrum of the current frame; or the ratio of the generated current frame's high-band spectrum energy to the low-band spectrum energy multiplied by a decimal, so that The value of the correction factor is less than 1, so that the high frequency spectrum energy value of the current frame is less than a certain multiple of its low frequency spectrum energy value, optionally, the high frequency spectrum energy value of the current frame is less than its low frequency spectrum energy value 0.8 times the value.

Step S211, converting the current frame from a frequency domain signal to a time domain signal. Here, the current frame includes its low-frequency time-domain signal and high-frequency time-domain signal.

Step S212, using the attenuation factor of the current frame to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame.

In this embodiment, the attenuation factor of the current frame is determined according to the frame type information of the current frame, the frame type information of the previous frame relative to the current frame and its attenuation factor; for each frame after the current frame, the The attenuation factor of a frame is determined according to the frame type information of the frame, the frame type information of a previous frame relative to the frame and its attenuation factor. During implementation, the determined attenuation factor of the current frame is multiplied by the high-frequency time-domain signal of the time-domain signal of the current frame. Wherein, the frame type information is a wideband frame or a narrowband frame. The specific implementation mode is shown in Fig. 3 for details.

Step S213, outputting the time domain signal of the current frame.

The code form of the above description is:

if((prev_mode==0)&&(E _gain1 <E _gain2 )&&(α*E _{c2_L} <E _{p1_L} ))α≥1.6

mode=1;

if((prev_mode==1)&&(δ*E _{p1_L} <E _{c2_L} ))δ≥1.6

mode=0;

For example: when the mode is set to 1 in the i-th frame, it is judged that the high-band spectrum of the i-th frame needs to be corrected, and the mode value is assigned to prev_mode at the same time; then the mode value of the i+1th frame is judged, If the mode is still set to 1, do the same operation as the i-th frame; then judge the mode value of the i+2, i+3... frame until the mode of the i+N-th frame is set to 0 , it is determined that it is not necessary to correct the high-band spectrum of the i+Nth frame.

This embodiment avoids the discontinuity of the filter state caused by the switching between speech signals of different bandwidths, resulting in an energy shock The phenomenon makes the decoded continuous speech signal more comfortable in hearing, and improves the subjective sound quality and auditory quality.

Fig. 3 is a schematic flowchart of determining an attenuation factor of a current frame in a speech decoding method according to an embodiment of the present invention. In Embodiment 2 or 3 of the present invention, optionally, a schematic flowchart of attenuating the high-frequency time-domain signal of the time-domain signal of the current frame by using the attenuation factor of the current frame. In the embodiment of the present invention, the attenuation factor is denoted as β, and its initial value is 1. As shown in Figure 3, determining the attenuation factor of the current frame includes the following steps:

Step S301, judging whether the current frame is a wideband frame or a narrowband frame. If the current frame is a broadband frame, execute step S302. If the current frame is a narrowband frame, execute step S303.

Step S302, the attenuation factor of the current frame is 1.

Step S303 , judging whether the original frame relative to the previous frame of the current frame is a wideband frame or a narrowband frame. If the original frame of the previous frame is a narrowband frame, step S304 is performed. If the original frame of the previous frame is a broadband frame, step S305 is executed.

Step S304 , the attenuation factor of the current frame is the square of the attenuation factor of the previous frame, that is, β _{current frame} = β _{previous frame} ² .

Step S305, the attenuation factor of the current frame is the value obtained by multiplying its initial value by a decimal γ. In this embodiment, 0.5<γ<0.7.

Optionally, determining the attenuation factor of the current frame can be performed after judging the frame type of the current frame and the original frame type of the previous frame relative to the current frame, which can be performed in step S204 to step S211, or step The process from S404 to step S412 is executed in parallel. Optionally, it may also be performed after step S211 or step S412 is performed. After the attenuation factor of the current frame is determined, step S212 or step S413 is executed to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame by using the determined attenuation factor of the current frame. Specifically, the determined attenuation factor of the current frame is multiplied by the high-frequency time-domain signal of the time-domain signal of the current frame.

Step S306: Multiply the high-frequency time-domain signal of the time-domain signal of the current frame by the attenuation factor of the current frame.

In the embodiment of the present invention, since when the high-frequency spectrum energy of blind expansion is larger than the actual high-frequency spectrum energy, it will have a great impact on the quality of the voice signal, therefore, when the actual high-frequency spectrum energy is not known In order to reduce the spectrum energy, by attenuating the high-band time-domain signal of the current frame, the blind-spread high-frequency spectrum energy should be reduced as much as possible, so that the quality of the decoded speech signal can be improved.

FIG. 4 is a schematic flowchart of a voice decoding method according to Embodiment 3 of the present invention. As shown in Figure 4, the continuous speech signal that receives is decoded, and this method comprises:

Step S401, judging whether the current frame is a wideband frame or a narrowband frame. If the current frame is a broadband frame, execute step S411. If the current frame is a narrowband frame, execute step S402.

Step S402, directly copy the low-band spectrum of the current frame to the high-frequency, generate the high-band spectrum of the current frame, and expand the current frame from a narrow-band frame to a wide-band frame.

Step S403 , judging whether the original frame relative to the previous frame of the current frame is a wideband frame or a narrowband frame. If the original frame of the previous frame is a broadband frame, step S404 is executed and then step S406 is executed. If the original frame of the previous frame is a narrowband frame, step S405 is executed and then step S407 is executed.

Step S404. Calculate the low-band spectrum energy value and the high-band spectrum energy value of the previous frame according to the frequency band information of the previous frame, and calculate the low-band spectrum energy value of the current frame according to the low-band information of the current frame.

In this embodiment, the energy of the low-band spectrum of the previous frame is marked as E _{p1_L} , the energy of the high-band spectrum is marked as E _{p1_H} , and the energy of the low-band spectrum of the current frame is marked as E _{c2_L} .

In this embodiment, the manner of calculating the low-band spectrum energy value and the high-band spectrum energy value of a frame signal is the same as the implementation manner described in step S103 of Embodiment 1, so details are not repeated here.

Step S405. Calculate the value of the low-band spectrum energy E _{p1_L} of the previous frame according to the low-band information of the previous frame, and calculate the value of the low-band spectrum energy E _{c2_L} of the current frame according to the low-band information of the current frame.

Step S406, according to the initial value of the mode bit of the previous frame, and the value of the low-band spectrum energy E _{p1_L} of the previous frame calculated in step S404, the value of the high-band spectrum energy E _{p1_H} , and the low-band spectrum of the current frame The value of the energy E _{c2_L} determines the value of the current frame mode bit.

In this embodiment, the previous frame mode bit is recorded as prev_mode with an initial value of 0, and the current frame mode bit is recorded as mode. The steps are specifically:

If the current frame satisfies the following three conditions at the same time, set mode to 1. The three conditions are:

1. prev_mode=0;

2. The energy E _{p1_H} of the low-frequency spectrum of the previous frame is greater than the energy E _{c2_L} of the low-frequency spectrum of the current frame by a certain multiple α. In this embodiment, α is generally above 1.6;

3. The energy E _{p1_H} of the high frequency spectrum of the previous frame is greater than the energy E _{c2_L} of the low frequency spectrum of the current frame.

Step S407, according to the value of the previous frame mode bit, and the value of the low-band spectral energy E _{p1_L} of the previous frame calculated in step S405, and the value of the low-band spectral energy E _{c2_L} of the current frame, determine the current frame mode bit value.

In this embodiment, the mode bit of the previous frame is recorded as prev_mode, and the mode bit of the current frame is recorded as mode. The steps are specifically:

If the frame meets the following two conditions at the same time, the mode position is set to 0. The two conditions are:

1. prev_mode=1;

2. The energy E _{c2_L} of the low-frequency spectrum of the current frame is greater than the energy E _{p1_H} of the low-frequency spectrum of the previous frame by a certain multiple α. In this embodiment, δ is generally above 1.6.

Step S408, judging whether the value of the mode bit of the current frame is 1. If the value is 1, it is determined that the current frame needs to be corrected, step S410 is executed and then step S412 is executed. If the value is 0, it is determined that the current frame does not need to be corrected, and step S411 is directly executed.

Step S409, assigning the value of the mode bit of the current frame to the mode bit of the previous frame to update the value of the mode bit of the previous frame in step S407.

Step S410, using a correction factor to correct the high frequency spectrum of the current frame, so that its energy is closer to the real high frequency spectrum energy.

For example, the high-band spectrum of the current frame is corrected by multiplying the high-band spectrum of the current frame by a correction factor.

In this embodiment, the value of the correction factor is specifically the average energy value of the low-band spectrum of the current frame; or the ratio of the generated current frame's high-band spectrum energy to the low-band spectrum energy multiplied by a decimal, so that The value of the correction factor is less than 1, so that the high frequency spectrum energy value of the current frame is less than a certain multiple of its low frequency spectrum energy value, optionally, the high frequency spectrum energy value of the current frame is less than its low frequency spectrum energy value 0.8 times the value.

Step S411: Multiply the high frequency band spectrum of the current frame by the high frequency frequency domain envelope of the previous frame relative to this frame.

Step S412, converting the current frame from a frequency domain signal to a time domain signal. Here, the current frame includes its low-frequency time-domain signal and high-frequency time-domain signal.

Step S413, using the attenuation factor of the current frame to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame.

In this embodiment, the implementation manner of step S413 is the same as that of step S212 in the second embodiment, and will not be repeated here.

Step S414, outputting the time domain signal of the current frame.

The code form of the above description is:

if((prev_mode==0)&&(E _{c2_L} <E _{p1_H} )&&(α*E _{c2_L} <E _{p1_H} ))α≥1.6

mode=1;

if((prev_mode==1)&&(δ*E _{p1_H} <E _{c2_L} ))δ≥1.6

mode=0;

For example: when the mode is set to 1 in the i-th frame, it is judged that the high-band spectrum of the i-th frame needs to be corrected, and the mode value is assigned to prev_mode at the same time; then the mode value of the i+1th frame is judged, If the mode is still set to 1, do the same operation as the i-th frame; then judge the mode value of the i+2, i+3... frame until the mode of the i+N-th frame is set to 0 , it is determined that it is not necessary to correct the high-band spectrum of the i+Nth frame.

This embodiment avoids the discontinuity of the filter state caused by the switching between speech signals of different bandwidths, resulting in an energy shock The phenomenon makes the decoded continuous speech signal more comfortable in hearing, and improves the subjective sound quality and auditory quality.

FIG. 5 is a schematic structural diagram of a speech decoding device according to Embodiment 4 of the present invention. As shown in FIG. 5 , the device includes: an expansion unit 10 , a first correction processing unit 11 , a transformation unit 12 , an attenuation unit 13 and an attenuation factor update unit 14 . The specific functions of each unit are described in detail below.

The extension unit 10 is used to multiply the normalized low-band spectrum of the current frame of the narrow-band by the high-band frequency domain envelope of the previous frame relative to the current frame and copy it to the high frequency to generate the current frame high frequency spectrum.

In the embodiment of the present invention, for each frame after the current frame that is narrowband, the extension unit 10 is also used to judge whether the frame is a narrowband frame, and if so, multiply the normalized low-band spectrum of the frame by Generates the high-band spectrum of the current frame with the high-band frequency-domain envelope of the previous frame relative to this frame copied to high frequencies.

The first correction processing unit 11 is used to judge whether the current frame expanded to wideband needs to be corrected according to the low frequency band information of the current frame and the frequency band information of the previous frame relative to the current frame, and when the judgment result is yes, use The correction factor corrects the high-frequency spectrum of the current frame, and when the determination result is negative, no correction is performed on the current frame.

Correspondingly, in the embodiment of the present invention, for all narrowband frames following the current narrowband frame, the first correction processing unit 11 is also configured to The frequency band information is used to judge whether the frame needs to be corrected. If the judgment result is yes, the high frequency band spectrum of the frame is corrected by the correction factor. When the judgment result is no, the frame is not corrected.

For example, in the embodiment of the present invention, the frame is a narrowband frame, and when the original frame relative to the previous frame of the frame is a broadband frame, the low-band information of the frame is the low-band spectrum energy value of the frame , and the high frequency band spectrum energy value of the frame generated according to the low frequency band spectrum of the frame, and the ratio of the high frequency band spectrum to the low frequency band spectrum energy of the frame, the frequency band information of the previous frame relative to the frame is the energy value of the low-band spectrum, the energy value of the high-band spectrum, and the energy ratio of the high-band spectrum to the low-band spectrum of the previous frame relative to the frame.

For example, in the embodiment of the present invention, the frame is a narrowband frame, and when the original frame relative to the previous frame of the frame is still a narrowband frame, the low-band information of the frame is the low-band spectral energy of the frame Value, the frequency band information of the previous frame relative to this frame is the low frequency band spectrum energy value of the previous frame relative to this frame.

FIG. 6 is a schematic structural diagram of the first modification processing unit 11 in the speech decoding device of FIG. 5 . As shown in FIG. 6 , the first correction processing unit 11 includes: a calculation unit 110 , a judgment unit 111 , and a judgment result processing unit 112 . The specific functions of each unit are described in detail below.

Calculation unit 110, for calculating the low-band spectrum energy value of the previous frame, the high-band spectrum energy value, the energy ratio of the high-band spectrum and the low-band spectrum of the previous frame according to the frequency band information of the previous frame, and It is used to calculate the low frequency band spectrum energy value of the current frame according to the low frequency band information of the current frame, and is used to calculate the high frequency band spectrum energy value of the current frame according to the generated high frequency band information of the current frame, and calculate the current frame The energy ratio of the high frequency band spectrum to the low frequency band spectrum.

Wherein, the manner of calculating the low-band spectral energy value, the high-frequency spectral energy value of a frame signal, and the ratio of the high-frequency spectral energy to the low-frequency spectral energy of a frame signal is detailed in the description of the method in the embodiment of the present invention , which will not be repeated here.

The judging unit 111 is used to calculate the energy value of the low frequency band spectrum of the previous frame calculated by the calculation unit 110 according to the value of the mode bit of the previous frame, and the energy ratio of the high frequency band spectrum and the low frequency band spectrum of the previous frame, and the current The energy value of the low frequency band spectrum of the frame, and the energy ratio of the high frequency band spectrum and the low frequency band spectrum of the current frame determine the value of the mode bit of the current frame; and judge whether the current frame needs to be corrected according to the value of the mode bit of the current frame.

Correspondingly, in this embodiment of the present invention, for all narrowband frames following the current frame that is a narrowband, the judging unit 111 is further configured to:

For example, this frame is a narrowband frame, when the original frame relative to the previous frame of this frame is a wideband frame, then according to the value of the mode bit of the previous frame, the calculation unit 110 calculates the previous frame relative to this frame The energy value of the low-band spectrum, the energy ratio of the high-band spectrum to the low-band spectrum, and the low-band spectrum energy value of the frame, the energy ratio of the high-band spectrum to the low-band spectrum, determine the value of the mode bit of the current frame; and Determine whether the frame needs to be corrected according to the value of the current frame mode bit;

For example, this frame is a narrowband frame, when the original frame relative to the previous frame of this frame is still a narrowband frame, then according to the value of the mode bit of the previous frame, the calculation unit 110 calculates the previous frame relative to this frame The low-band spectral energy value of the frame, and the low-frequency spectral energy value of the frame determine the value of the mode bit of the current frame; and determine whether the frame needs to be corrected according to the value of the mode bit of the current frame.

The judging result processing unit 112 is configured to correct the high frequency band spectrum of the current frame by using the correction factor when the judging result is yes, and not to modify the current frame when the judging result is no.

The transformation unit 12 is configured to transform the current frame processed according to the judgment result from a signal in the frequency domain to a signal in the time domain.

The attenuation unit 13 is configured to use the attenuation factor of the current frame to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame, and then output the time-domain signal of the current frame.

Optionally, in the embodiment of the present invention, the device further includes: an attenuation factor update unit 14, configured to use the frame type information of the current frame, the frame type information of the previous frame relative to the current frame, and the value of the attenuation factor , to determine the current value of the attenuation factor. This value is used by the attenuation unit 13 to attenuate the high frequency time domain signal. The frame type information is wideband frame or narrowband frame. For the updating process of the attenuation factor updating unit 14, refer to the description of the method in the embodiment of the present invention for details, and details are not repeated here.

FIG. 7 is a schematic structural diagram of a speech decoding device according to Embodiment 5 of the present invention. As shown in FIG. 7 , the device includes: an expansion unit 10 ′, a first correction processing unit 11 ′, a second correction processing unit 15 , a transformation unit 12 , an attenuation unit 13 and an attenuation factor update unit 14 . The specific functions of each unit are described in detail below.

The extension unit 10' is configured to directly copy the low-band spectrum of the current frame for the narrow-band to the high frequency, so as to generate the high-band spectrum of the current frame.

In the embodiment of the present invention, for each frame after the current frame that is a narrowband, the extension unit 10 is also used to judge whether the frame is a narrowband frame, and if so, copy the low-band spectrum of the frame to the high-frequency, Generate a high-band spectrum for the current frame.

The first correction processing unit 11' is used to judge whether the current frame extended to wideband needs to be corrected according to the frequency band information of the current frame and the low frequency band information of the previous frame relative to the current frame, and when the judgment result is yes, The high-frequency spectrum of the current frame is corrected by using the correction factor, and when the determination result is negative, no correction is performed on the current frame.

Correspondingly, in the embodiment of the present invention, for all narrowband frames following the current narrowband frame, the first modification processing unit 11' is also used to The frequency band information of the frame is used to judge whether the frame needs to be corrected. If the judgment result is yes, the high frequency band spectrum of the frame is corrected by the correction factor. When the judgment result is no, the frame is not corrected.

For example, in the embodiment of the present invention, the frame is a narrowband frame, and when the original frame relative to the previous frame of the frame is a broadband frame, the low-band information of the frame is the low-band spectrum energy value of the frame , and the high frequency band spectrum energy value of the frame generated according to the low frequency band spectrum of the frame, the frequency band information of the previous frame relative to the frame is the low frequency band spectrum energy value of the previous frame relative to the frame and High frequency band spectral energy value.

For example, in the embodiment of the present invention, the frame is a narrowband frame, and when the original frame relative to the previous frame of the frame is still a narrowband frame, the low-band information of the frame is the low-band spectral energy of the frame Value, the frequency band information of the previous frame relative to this frame is the low frequency band spectrum energy value of the previous frame relative to this frame.

In the embodiment of the present invention, the first correction processing unit 11' includes: a calculation unit 110', a judgment unit 111', and a judgment result processing unit 112'. The specific functions of each unit are described in detail below.

Calculation unit 110', used to calculate the low-band spectrum energy value and high-band spectrum energy value of the previous frame according to the frequency band information of the previous frame, and to calculate the low-band spectrum energy value of the current frame according to the low-frequency band information of the current frame value.

The judging unit 111' is used to calculate the low-band spectrum energy value and the high-band spectrum energy value of the previous frame calculated by the calculation unit 110' according to the value of the mode bit of the previous frame, and the low-band spectrum energy value of the current frame, determining the value of the mode bit of the current frame; and judging whether the current frame needs to be corrected according to the value of the mode bit of the current frame.

Correspondingly, in the embodiment of the present invention, for all narrowband frames following the current frame that is a narrowband, the judging unit 111' is further configured to:

For example, this frame is a narrowband frame, when the original frame relative to the previous frame of this frame is a wideband frame, then according to the value of the mode bit of the previous frame, the calculation unit 110' calculates the previous frame relative to the frame The low-band spectral energy value and the high-band spectral energy value of the frame, and the low-frequency spectral energy value of the frame, determine the value of the current frame mode bit; and judge whether the frame needs to be corrected according to the value of the current frame mode bit;

For example, this frame is a narrowband frame, when the original frame relative to the previous frame of this frame is still a narrowband frame, then according to the value of the mode bit of the previous frame, the calculation unit 110' calculates the previous frame relative to this frame The low-band spectral energy value of a frame, and the low-frequency spectral energy value of the frame determine the value of the mode bit of the current frame; and determine whether the frame needs to be corrected according to the value of the mode bit of the current frame.

The judging result processing unit 112' is configured to correct the high frequency band spectrum of the current frame by using the correction factor when the judging result is yes, and not to modify the current frame when the judging result is no.

The second correction processing unit 15 is configured to multiply the high frequency frequency spectrum of the current frame by For the high-band frequency domain envelope, when the first correction processing unit 11 ′ has corrected the high-frequency spectrum of the current frame, it does not correct the current frame.

Correspondingly, in the embodiment of the present invention, for all narrowband frames following the current frame that is a narrowband, the second correction processing unit 15 is also used to When performing correction, the high frequency band spectrum of the frame is multiplied by the high frequency band frequency domain envelope of the previous frame relative to the frame, when the first correction processing unit 11' has corrected the high frequency band spectrum of the frame, The frame is processed without correction.

The transformation unit 12 is configured to transform the current frame processed according to the judgment result from a signal in the frequency domain to a signal in the time domain.

The attenuation unit 13 is configured to use the attenuation factor of the current frame to attenuate the high-frequency time-domain signal of the time-domain signal of the current frame, and then output the time-domain signal of the current frame.

Optionally, in the embodiment of the present invention, the device further includes: an attenuation factor update unit 14, configured to use the frame type information of the current frame, the frame type information of the previous frame relative to the current frame, and the value of the attenuation factor , to determine the current value of the attenuation factor, which is used by the attenuation unit 13 to attenuate the high-frequency time-domain signal. The frame type information is wideband frame or narrowband frame. Similarly, for the updating process of the attenuation factor updating unit 14, refer to the description of the method in the embodiment of the present invention for details, and details are not repeated here.

As can be seen from the above embodiments, according to a speech decoding method and device proposed in the embodiments of the present invention, the low-band spectrum of the narrow-band frame is used to generate the high-band spectrum of the narrow-band frame, and the narrow-band frequency spectrum is judged according to certain judgment conditions. Whether the frame needs to be corrected, when it needs to be corrected, correct the high-frequency spectrum of the narrow-band frame, so that the narrow-band frame in the voice signal is expanded into an effective wide-band frame, which solves the problem of hearing loss caused by voice signals with different bandwidths. Improve the auditory quality of the speech signal.

At the same time, the application of the present invention in the frequency band extension technique of expanding the narrow-band frame into a wide-band frame can effectively improve the performance of the recovered high-band signal when the low-band energy is low.

The above disclosure is only a preferred embodiment of the present invention, which certainly cannot limit the scope of rights of the present invention. Therefore, equivalent changes made according to the claims of the present invention still fall within the scope of the present invention.

Claims (14)

1. tone decoding method comprises:

The low-frequency band frequency spectrum that is utilized as the present frame of narrow-band generates the high frequency band frequency spectrum of described present frame;

According to the low-frequency band information of described present frame with respect to the band information of the former frame of described present frame, judge whether described present frame needs to revise, if judged result is for being, then utilize modifying factor that the high frequency band frequency spectrum of described present frame is revised, if judged result is not then revised described present frame for not;

Described present frame after will handling according to described judged result is transformed to time-domain signal from frequency-region signal;

After utilizing the decay factor of described present frame that the high frequency time-domain signal of described present frame time-domain signal is decayed, export the time-domain signal of described present frame.

2. the method for claim 1 is characterized in that, described former frame with respect to described present frame is the broadband frame.

3. method as claimed in claim 2, it is characterized in that, the high frequency band frequency spectrum that the described low-frequency band frequency spectrum that is utilized as the present frame of narrow-band generates described present frame is specially: the high frequency band frequency domain envelope that the normalized low-frequency band frequency spectrum of described present frame be multiply by with respect to the former frame of described present frame copies to high frequency, generates the high frequency band frequency spectrum of described present frame;

Described according to described present frame low-frequency band information and with respect to the band information of the former frame of described present frame, judge whether described present frame needs to revise, and comprising:

The low-frequency band spectrum energy value and the high frequency band spectrum energy value of the described former frame that calculates with the value of former frame pattern position, according to the band information of described former frame, and the low-frequency band spectrum energy value and the high frequency band spectrum energy value of the described present frame that goes out according to the low-frequency band information calculations of described present frame, determine the value of present frame pattern position;

Judge the whether needs correction of described present frame according to the value of described present frame pattern position.

4. method according to claim 2, it is characterized in that, the high frequency band frequency spectrum that the described low-frequency band frequency spectrum that is utilized as the present frame of narrow-band generates described present frame is specially: the low-frequency band frequency spectrum of described present frame is directly copied to high frequency, generate the high frequency band frequency spectrum of described present frame;

Described according to described present frame low-frequency band information and with respect to the band information of the former frame of described present frame, judge whether described present frame needs to revise, and comprising:

The low-frequency band spectrum energy value of the described former frame that calculates with the value of former frame pattern position, according to the band information of described former frame and high frequency band spectrum energy value, and the low-frequency band spectrum energy value of the described present frame that goes out according to the low-frequency band information calculations of described present frame are determined the value of present frame pattern position;

Judge the whether needs correction of described present frame according to the value of described present frame pattern position.

5. method as claimed in claim 4 is characterized in that, the present frame after described will the processing according to described judged result is transformed to time-domain signal from frequency-region signal, also comprises:

If the high frequency band frequency spectrum of described present frame is not revised according to described judged result, then the normalized high frequency band frequency spectrum of present frame be multiply by the high frequency band frequency domain envelope of described former frame after, this present frame is transformed to time-domain signal from frequency-region signal.

6. method as claimed in claim 2 is characterized in that, described method further comprises:

For described be each frame behind the present frame of narrow-band, judge whether this frame is the arrowband frame, if utilize the low-frequency band frequency spectrum of described this frame to generate the high frequency band frequency spectrum of described this frame;

According to the low-frequency band information of described this frame with respect to the band information of the former frame of described this frame, judge whether described this frame needs to revise, if judged result is for being, then utilize modifying factor that the high frequency band frequency spectrum of described this frame is revised, if judged result is not then revised described this frame for not;

Described this frame after will handling according to judged result is transformed to time-domain signal from frequency-region signal;

After utilizing the decay factor of described this frame that the high frequency time-domain signal of described this frame time-domain signal is decayed, the time-domain signal of described this frame of output.

7. method as claimed in claim 6 is characterized in that the decay factor of described present frame is determined according to the frame type information of the former frame of the frame type information of described present frame, described present frame; With

The decay factor of described this frame determines that according to the frame type information and the decay factor thereof of the frame type information of described this frame, described former frame with respect to this frame described frame type information is broadband frame or narrow-band frame.

8. method according to claim 1 is characterized in that, the value of described modifying factor is determined according to the low-frequency band information of described present frame.

9. an audio decoding apparatus is characterized in that, comprising:

Expanding element, the low-frequency band frequency spectrum that is used to be utilized as the present frame of narrow-band generates the high frequency band frequency spectrum of described present frame;

The first correcting process unit, be used for according to the low-frequency band information of described present frame with respect to the band information of the former frame of described present frame, judge whether described present frame needs to revise, when judged result when being, utilize modifying factor that the high frequency band frequency spectrum of described present frame is revised, judged result is not revised described present frame for not the time;

Converter unit is used for the present frame after handling according to described judged result is transformed to time-domain signal from frequency-region signal;

Attenuation units, be used to utilize the decay factor of described present frame that the high frequency time-domain signal of described present frame time-domain signal is decayed after, export the time-domain signal of described present frame.

10. device as claimed in claim 9, it is characterized in that, described expanding element is used for the high frequency band frequency domain envelope that normalized low-frequency band frequency spectrum with described present frame multiply by with respect to the former frame of described present frame and copies to high frequency, generates the high frequency band frequency spectrum of described present frame;

The described first correcting process unit comprises:

Computing unit, be used for calculating the low-frequency band spectrum energy value and the high frequency band spectrum energy value of described former frame, and be used for going out the low-frequency band spectrum energy value and the high frequency band spectrum energy value of described present frame according to the low-frequency band information calculations of described present frame according to the band information of described former frame;

Judging unit is used for low-frequency band spectrum energy value and high frequency band spectrum energy value according to the value of former frame pattern position, described former frame, and the low-frequency band spectrum energy value and the high frequency band spectrum energy value of described present frame, determines the value of present frame pattern position; With judge the whether needs correction of described present frame according to the value of described present frame pattern position;

The judged result processing unit, be used for when judged result when being, utilize modifying factor that the high frequency band frequency spectrum of described present frame is revised, when judged result for not the time, described present frame is not revised.

11. device as claimed in claim 9 is characterized in that, described expanding element is used for the low-frequency band frequency spectrum of described present frame is directly copied to high frequency, generates the high frequency band frequency spectrum of described present frame;

The described first correcting process unit comprises:

Computing unit is used for calculating according to the band information of described former frame the low-frequency band spectrum energy value and the high frequency band spectrum energy value of described former frame, and is used for going out according to the low-frequency band information calculations of described present frame the low-frequency band spectrum energy value of described present frame;

Judging unit is used for low-frequency band spectrum energy value and high frequency band spectrum energy value according to the value of former frame pattern position, described former frame, and the low-frequency band spectrum energy value of described present frame, determines the value of present frame pattern position; With judge the whether needs correction of described present frame according to the value of described present frame pattern position;

The judged result processing unit, be used for when judged result when being, utilize modifying factor that the high frequency band frequency spectrum of described present frame is revised, when judged result for not the time, described present frame is not revised.

12. device as claimed in claim 11 is characterized in that, described device also comprises:

The second correcting process unit, be used for when the described first correcting process unit is not revised the high frequency band frequency spectrum of described present frame, the normalized high frequency band frequency spectrum of described present frame be multiply by the high frequency band frequency domain envelope of described former frame, when the described first correcting process unit is out-of-date to the high frequency band frequency spectrum correction of described present frame, described present frame is done the processing of not revising;

Described converter unit also is used for the described present frame after carrying out correcting process through the described second correcting process unit is transformed to time-domain signal from frequency-region signal.

13. device as claimed in claim 9 is characterized in that, described expanding element, also be used for, for described be each frame behind the present frame of narrow-band, judge whether this frame is the arrowband frame, if generate the high frequency band frequency spectrum of described this frame according to the low-frequency band frequency spectrum of described this frame;

The described first correcting process unit, also be used for according to the low-frequency band information of described this frame with respect to the band information of the former frame of described this frame, judge whether described this frame needs to revise, when judged result when being, utilize modifying factor that the high frequency band frequency spectrum of described this frame is revised, when judged result for not the time, described this frame is not revised.

14. device as claimed in claim 13 is characterized in that, described device also comprises:

The decay factor updating block, be used for according to described present frame frame type information, determine the decay factor of described present frame with respect to the frame type information of the former frame of described present frame; Perhaps

According to the frame type information of described this frame, with respect to the frame type information of the former frame of described this frame and the decay factor that decay factor is determined described this frame thereof, described frame type information is broadband frame or narrow-band frame.

Images