CN114333870A - Voice processing method and device - Google Patents

Abstract

The present invention discloses a voice processing method and a related device, wherein the method comprises: performing echo cancellation on n channels of first voice signals to obtain n channels of second voice signals, wherein the first voice signals are obtained by collecting voice signals; perform beamforming on the n channels of second voice signals to obtain m channels of first beams; obtain interference candidate beams from the m channels of first beams; The second voice signal is processed to obtain the target voice signal, which can improve the signal quality of the target voice signal.

Description

Voice processing method and device

technical field

The present invention relates to the field of speech processing, and in particular, to a speech processing method and device.

Background technique

In the prior art solution, when voice recognition and voice wake-up are performed, the method of voice enhancement of the received voice signal is usually used to improve the success rate of recognition or voice wake-up. However, in harsh scenarios (such as strong external noise scenarios), Some speech enhancement methods usually cannot enhance the signal well, resulting in a low success rate in speech recognition or speech wake-up.

SUMMARY OF THE INVENTION

Embodiments of the present invention provide a voice processing method and device, which can perform echo cancellation on multi-channel voice signals, and process the echo-cancelled voice signals according to beams obtained by beamforming the echo-cancelled voice signals to obtain a target voice signal, The signal quality of the target voice signal can be improved.

In a first aspect, an embodiment of the present invention provides a speech processing method, the method includes:

Performing echo cancellation on n-channel first voice signals to obtain n-channel second voice signals, where the first voice signals are collected voice signals;

performing beamforming on the n channels of second speech signals to obtain m channels of first beams;

obtaining interference candidate beams from the m first beams;

The n channels of second speech signals are processed according to the interference candidate beams to obtain a target speech signal.

In this example, echo cancellation is performed on the first speech signal to obtain the second speech signal, and the second speech signal is processed according to the interference candidate beams formed by the beamforming of the second speech signal to obtain the target speech signal. Therefore, the target speech signal can be obtained by The interference candidate beam processes the second speech signal to obtain the target speech signal, which can improve the signal quality of the target speech signal.

With reference to the first aspect, in a possible implementation manner, the obtaining interference candidate beams from the m-channel first beams includes:

obtaining the frame-level energy corresponding to the m first beams to obtain m first frame-level energy values;

determining the average energy value of the beam according to the m first frame-level energy values;

determining a count value corresponding to the m first beams according to the m first frame-level energy values and the beam average energy value;

The beam corresponding to the largest count value among the count values corresponding to the m first beams is determined as the interference candidate beam.

In this example, the first beams are counted according to the frame-level energy corresponding to each first beam and the average energy of the beam, and the beam corresponding to the maximum count is determined as the interference candidate beam. The interference candidate beam is obtained from the angle, which improves the accuracy of the interference candidate acquisition.

With reference to the first aspect, in a possible implementation manner, the processing of the n channels of second speech signals according to the interference candidate beams to obtain the target speech signal includes:

obtaining the smoothed energy value of the interference candidate beam;

determining a filter strength value according to the smoothed energy value;

Perform filtering processing on the n channels of second speech signals according to the filter strength value and the interference candidate beam to obtain n channels of third speech signals;

The n third voice signals are processed to obtain a target voice signal.

In this example, the second voice signal is filtered through the filtering strength value determined by the smoothing energy and the interference beam to obtain the third filtered signal, and the third filtered signal is processed to obtain the target voice signal. Filtering the second speech by the beam can improve the quality of the filtered third signal, thereby improving the signal quality of the target speech signal.

With reference to the first aspect, in a possible implementation manner, the processing of the n third voice signals to obtain the target voice signal includes:

Obtain the ith third voice signal and the jth third voice signal from the n third voice signals;

Perform de-reverberation and blind source separation on the i-th third speech signal and the j-th third speech signal to obtain the first candidate speech signal and the second candidate speech signal;

performing beamforming on the n channels of third speech signals to obtain m channels of second beams;

obtaining a first target beam from the m-way second beam;

performing at least noise reduction processing on the first target beam to obtain a processed first target beam;

The target speech signal is determined according to the first candidate speech signal and the second candidate speech signal and the processed first target beam.

In this example, de-reverberation and blind source separation are performed on the i-th third speech signal and the j-th third speech signal to obtain the first candidate speech signal and the second candidate speech signal, and according to the first candidate speech signal The signal, the second candidate speech signal, and the noise-reduced first target beam are used to determine the target speech signal, which can improve the signal quality of the target speech signal.

With reference to the first aspect, in a possible implementation manner, the processing of the n channels of second speech signals according to the interference candidate beams to obtain the target speech signal includes:

Determine whether the interference candidate beam is a preset beam, and if the interference candidate beam is a preset beam, perform filtering processing on the n channels of second speech signals according to the interference candidate beam and the preset filtering strength to obtain n fourth voice signal;

performing beamforming on the n channels of fourth speech signals to obtain m channels of third beams;

obtaining a second target beam from the m third beams;

performing at least noise reduction processing on the second target beam to obtain a processed second target beam;

Obtain the h-th second voice signal and the k-th second voice signal from the n-way second voice signals;

Filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal to obtain a third candidate voice signal and a fourth candidate voice signal;

The target speech signal is determined according to the third candidate speech signal and the fourth candidate speech signal and the processed second target beam.

In this example, after it is determined that the interference candidate beam is the preset beam, the second speech signal is filtered according to the interference candidate beam and the preset filter strength, and the second target beam is at least subjected to noise reduction processing and the first The second voice signal of the h channel and the second voice signal of the kth channel are subjected to de-reverberation and blind source separation, and the third candidate voice signal and the fourth candidate voice signal are obtained to determine the target signal, which can improve the target voice signal signal quality.

With reference to the first aspect, in a possible implementation manner, the method further includes:

If the interference candidate beam is not a preset beam, the target speech signal is determined according to the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal.

If the interference candidate beam is not a preset beam, the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal are determined to determine the target speech signal, which can reduce resource consumption when determining the target speech signal.

In a second aspect, an embodiment of the present application provides a voice processing apparatus, and the apparatus includes:

an elimination unit, configured to perform echo cancellation on n-channel first voice signals to obtain n-channel second voice signals, where the first voice signals are collected voice signals;

a beamforming unit, configured to perform beamforming on the n channels of second speech signals to obtain m channels of first beams;

an obtaining unit, configured to obtain interference candidate beams from the m first beams;

The processing unit is configured to process the n channels of second speech signals according to the interference candidate beams to obtain a target speech signal.

With reference to the second aspect, in a possible implementation manner, the obtaining unit is specifically configured to:

obtaining the frame-level energy corresponding to the m first beams to obtain m first frame-level energy values;

determining the average energy value of the beam according to the m first frame-level energy values;

determining a count value corresponding to the m first beams according to the m first frame-level energy values and the beam average energy value;

The beam corresponding to the largest count value among the count values corresponding to the m first beams is determined as the interference candidate beam.

With reference to the second aspect, in a possible implementation manner, the processing unit is specifically configured to:

obtaining the smoothed energy value of the interference candidate beam;

determining a filter strength value according to the smoothed energy value;

Perform filtering processing on the n channels of second speech signals according to the filter strength value and the interference candidate beam to obtain n channels of third speech signals;

The n third voice signals are processed to obtain a target voice signal.

With reference to the second aspect, in a possible implementation manner, in the aspect of processing the n third voice signals to obtain the target voice signal, the processing unit is specifically configured to:

Obtain the ith third voice signal and the jth third voice signal from the n third voice signals;

Perform de-reverberation and blind source separation on the i-th third speech signal and the j-th third speech signal to obtain the first candidate speech signal and the second candidate speech signal;

performing beamforming on the n channels of third speech signals to obtain m channels of second beams;

obtaining a first target beam from the m-way second beam;

performing at least noise reduction processing on the first target beam to obtain a processed first target beam;

The target speech signal is determined according to the first candidate speech signal and the second candidate speech signal and the processed first target beam.

With reference to the second aspect, in a possible implementation manner, the processing unit is configured to:

Determine whether the interference candidate beam is a preset beam, and if the interference candidate beam is a preset beam, perform filtering processing on the n channels of second speech signals according to the interference candidate beam and the preset filtering strength to obtain n fourth voice signal;

performing beamforming on the n channels of fourth speech signals to obtain m channels of third beams;

obtaining a second target beam from the m third beams;

performing at least noise reduction processing on the second target beam to obtain a processed second target beam;

Obtain the h-th second voice signal and the k-th second voice signal from the n-way second voice signals;

Filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal to obtain a third candidate voice signal and a fourth candidate voice signal;

The target speech signal is determined according to the third candidate speech signal and the fourth candidate speech signal and the processed second target beam.

With reference to the second aspect, in a possible implementation manner, the processing unit is further configured to:

If the interference candidate beam is not a preset beam, the target speech signal is determined according to the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal.

In a third aspect, an embodiment of the present invention provides a voice processing apparatus, including:

memory for storing instructions; and

at least one processor coupled to the memory;

Wherein, when the at least one processor executes the instructions, the instructions cause the processor to execute all or part of the method shown in the first aspect.

In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the program instructions cause the processing when executed by a processor The processor executes all or part of the method shown in the first aspect.

These and other aspects of the invention will be more clearly understood from the description of the following embodiments.

Description of drawings

In order to explain the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

FIG. 1 provides a schematic diagram of a speech enhancement in an embodiment of the present application;

FIG. 2 provides a schematic flowchart of a voice processing method according to an embodiment of the present application;

FIG. 3 is a schematic structural diagram of a voice processing apparatus according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of another voice processing apparatus according to an embodiment of the present invention.

Detailed ways

The embodiments of the present application will be described below with reference to the accompanying drawings.

The following first introduces the main methods of introducing voice enhancement for voice wake-up and voice enhancement. As shown in Figure 1, voice enhancement is performed on the input signal, and voice wake-up or voice recognition is performed through the enhanced voice signal. The application scenarios of wake-up and voice recognition can be human-computer voice interaction scenarios. In human-computer voice interaction scenarios, more and more people use voice interaction to wake up electronic devices through voice or send instructions to electronic devices through voice.

The electronic device of the present application can be a terminal device such as a mobile phone, a tablet, a speaker, a TV, etc., for example, as a mobile phone, a tablet, a speaker, a TV, etc., the voice processing method can be to process the signal with interference picked up by the microphone, and output a clean Speech signal, as input to wake-up and recognition engine.

The process of the speech processing method is described in detail below.

Referring to FIG. 2, FIG. 2 provides a schematic flowchart of a speech processing method according to an embodiment of the present application. As shown in Figure 2, the speech processing method includes:

S201. Perform echo cancellation on n channels of first voice signals to obtain n channels of second voice signals, where the first voice signals are collected voice signals.

The n-channel first voice signals may be n-channel voice signals obtained through a microphone array/sound pickup array, and the microphone array may include n microphones and the like. n is a positive integer greater than or equal to 2.

The method for performing echo cancellation on the n-channel first voice signals may be to perform echo cancellation on the n-channel first voice signals according to a reference signal, specifically, performing echo cancellation by using an ACE algorithm, and the reference signal may be a preset signal. .

S202. Perform beamforming on n channels of second speech signals to obtain m channels of first beams.

A beamforming algorithm or the like may be used to perform beamforming on the n channels of second speech signals to obtain m channels of first beams. m can be a value independent of n, for example, m can be the same as n, or m can be different from n.

S203. Obtain interference candidate beams from the m first beams.

The interference candidate beams may be determined by the frame-level energy values corresponding to the m first beams. Specifically, the interference candidate beam may be determined according to the frame-level energy value and the beam average energy value corresponding to the m first beams. The average beam energy value can be understood as the average beam energy of m beams.

S204. Process the n channels of second speech signals according to the interference candidate beams to obtain a target speech signal.

The n channels of second speech signals may be filtered to obtain n channels of third speech signals after filtering, and the target speech signals may be obtained by processing according to the n channels of third speech signals. Or, it is judged whether the interference candidate beam is a preset beam, and the second voice signal is processed according to the judgment result to obtain the target voice signal. The preset beam can be a continuous interference beam, and the continuous interference beam can be understood as: there is continuous noise interference in the beam.

In this example, echo cancellation is performed on the first speech signal to obtain the second speech signal, and the second speech signal is processed according to the interference candidate beams formed by the beamforming of the second speech signal to obtain the target speech signal. Therefore, the target speech signal can be obtained by The interference candidate beam processes the second speech signal to obtain the target speech signal, which can improve the signal quality of the target speech signal.

In a possible implementation, a possible method for obtaining interference candidate beams from m-channel first beams includes:

A1. Acquire the frame-level energy corresponding to the m first beams to obtain m first frame-level energy values;

A2. Determine the average energy value of the beam according to the m first frame-level energy values;

A3. Determine the count value corresponding to the m first beams according to the m first frame-level energy values and the beam average energy value;

A4. Determine the beam corresponding to the largest count value among the count values corresponding to the m first beams as the interference candidate beam.

The frame-level energy value corresponding to the first beam may be obtained by the number of time domain sampling points of each frame of data corresponding to the first beam. Specifically, the frame-level energy can be obtained by the method shown in the following formula:

Among them, i is the ith first beam, the value range is [1, m], K represents the number of time domain sampling points of each frame of data, k value range [1, K], P _i is the ith first beam The frame-level energy of the beam.

The average value of the energy values of the m channels at the first frame level can be directly performed to obtain the average energy value of the beam. Specifically, the average energy value of the beam can be obtained by the method shown in the following formula:

Among them, P _avg is the average energy value of the beam, m is the amount of total frame energy, and P _i is the ith frame-level energy value.

According to the m first frame-level energy values and the beam average energy value, the method for determining the count value corresponding to the m first beams may specifically be:

First, the VAD module judges whether the current frame of each beam is in a quiet state according to the data of the beam data B1, B2, .

When the current frame of the i-th beam is in a non-quiet state, the counter C _i corresponding to the beam is updated as follows:

Among them, P _avg is the average energy value of the beam, and P _i is the ith first frame-level energy value.

In this example, the first beams are counted according to the frame-level energy corresponding to each first beam and the average energy of the beam, and the beam corresponding to the maximum count is determined as the interference candidate beam. The interference candidate beam is obtained from the angle, which improves the accuracy of the interference candidate acquisition.

In a possible implementation manner, a possible method for processing the n channels of second speech signals according to the interference candidate beams to obtain a target speech signal includes:

B1. Obtain the smoothed energy value of the interference candidate beam;

B2. Determine the filter strength value according to the smoothed energy value;

B3. Perform filtering processing on the n-channel second speech signals according to the filtering strength value and the interference candidate beam to obtain n-channel third speech signals;

B4. Process the n third voice signals to obtain a target voice signal.

The smoothed energy value can be determined by the smoothed energy value of the previous frame of the data frame and the maximum energy value of the beam of the current data frame. Specifically, the smoothed energy value can be determined by the method shown in the following formula:

为干扰候选波束的平滑能量值，

为上一帧的平滑能量值，P_max为本数据帧中的波束的最大能量值。in,

is the smoothed energy value of the interfering candidate beam,

is the smoothed energy value of the previous frame, and _Pmax is the maximum energy value of the beam in this data frame.

The filter strength value can be determined by the maximum beam energy, which can be determined by the method shown in the following formula:

超过高阈值H时，β＝1；当

低于低阈值L时，β＝0。H、L为预先设定的值，可以通过经验值或历史数据设定。Among them, β represents the filter strength, H represents the high threshold, L represents the low threshold, and H>L. when

When the high threshold H is exceeded, β=1; when

Below the low threshold L, β=0. H and L are preset values, which can be set through empirical values or historical data.

The interference candidate beam may be determined as the reference beam, and the filtering gain may be adjusted by the filtering strength value, so as to implement filtering processing on n channels of second speech signals to obtain n channels of third speech signals. When filtering the second speech signal, the filtering may be performed by a linear adaptive filtering method, for example, by Kalman filtering. Specifically, the interference candidate beam may be used as a reference beam, and the gain of the filter may be adjusted by the filtering strength value. to obtain the third voice signal.

Beamforming can be performed from n channels of third speech signals to obtain m channels of second beams, the processed first target beam can be determined according to the second beams, and two channels of speech signals can be obtained from n channels of third speech signals, and the two channels of speech signals can be obtained. The voice signal of the channel is processed to obtain a candidate voice signal, and the target voice signal is determined according to the candidate voice signal and the processed target beam.

In this example, the second voice signal is filtered through the filtering strength value determined by the smoothing energy and the interference beam to obtain the third filtered signal, and the third filtered signal is processed to obtain the target voice signal. Filtering the second speech by the beam can improve the quality of the filtered third signal, thereby improving the signal quality of the target speech signal.

In a possible implementation manner, a possible method for processing n channels of third voice signals to obtain a target voice signal includes:

C1, obtain the ith third voice signal and the jth third voice signal from the n third voice signals;

C2, perform de-reverberation and blind source separation on the i-th third voice signal and the j-th third voice signal to obtain the first candidate voice signal and the second candidate voice signal;

C3, performing beamforming on the n third voice signals to obtain m second beams;

C4. Obtain the first target beam from the m second beams;

C5. Perform at least noise reduction processing on the first target beam to obtain a processed first target beam;

C6. Determine the target speech signal according to the first candidate speech signal, the second candidate speech signal and the processed first target beam.

The method for acquiring the i-th third voice signal and the j-th third voice signal may be: determining the third voice signals corresponding to the two farthest microphones in the microphone array as the i-th third voice signal and the j-th third voice signal The third voice signal of the road.

De-reverberation and blind source separation are performed on the i-th third voice signal and the j-th third voice signal, to obtain the first candidate voice signal and the second candidate voice signal, specifically, for the i-th third voice signal. The voice signal and the j-th third voice signal are subjected to cross operation, etc., to obtain the first candidate voice signal and the second candidate voice signal. The method for performing de-reverberation and blind source separation on the third speech signal may be to obtain the first candidate speech signal and the second candidate speech signal through a de-reverberation module and a blind source separation module, and the blind source separation module may pass IVA. Algorithm implementation.

For the method for obtaining the first target beam from the m-channel second beam, reference may be made to the method for obtaining the interference candidate beam in the foregoing embodiment, which will not be repeated here.

According to the first candidate speech signal, the second candidate speech signal and the processed first target beam, the method for determining the target speech signal may be as follows: Any one of the processed first target beams is determined as the target speech signal, or any combination of the first candidate speech signal, the second candidate speech signal and the processed first target beam can be determined as the target The speech signal may also be any combination of the first candidate speech signal, the second candidate speech signal and the processed first target beam, and the processed signal is determined as the target speech signal.

The first target beam path wake-up enhancement signal effectively improves the effect of speech enhancement and interference suppression in external noise scenarios by filtering the interference beam. The quality of the target signal extraction in the same orientation scene is improved, and the quality of the target voice signal is improved according to the processed first beam, the first voice signal and the second voice signal, so that the target voice signal is used for voice recognition and voice wake-up. , which can improve the accuracy.

In a possible implementation manner, another possible method for processing the n channels of second speech signals according to the interference candidate beams to obtain the target speech signal may be:

D1. Determine whether the interference candidate beam is a preset beam, and if the interference candidate beam is a preset beam, filter the n second voice signals according to the interference candidate beam and a preset filtering strength, to obtain n fourth voice signals;

D2, performing beamforming on the n channels of fourth speech signals to obtain m channels of third beams;

D3. Obtain the second target beam from the m third beams;

D4. Perform at least noise reduction processing on the second target beam to obtain a processed second target beam;

D5, obtain the h-th second voice signal and the k-th second voice signal from the n-way second voice signal;

D6, filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal to obtain a third candidate voice signal and a fourth candidate voice signal;

D7. Determine the target speech signal according to the third candidate speech signal, the fourth candidate speech signal and the processed second target beam.

For the specific implementation of the above steps D2-D4, reference may be made to the method for obtaining the first target beam in the foregoing embodiment, and details are not described herein again.

Filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal, and demixing can also be performed on the h-th second voice signal and the k-th second voice signal sound to obtain the third candidate speech signal and the fourth candidate speech signal. For the method of filtering the h-th second voice signal and the k-th second voice signal, reference may be made to the foregoing method for filtering n-th second voice signals.

The preset beam can be a continuous interference beam, and the continuous interference beam can be understood as: there is continuous noise interference in the beam.

Obtaining the h-th channel of the second voice signal and the k-th channel of the second voice signal from the n-channel second voice signals can be performed by referring to the method for acquiring the i-th channel of the third voice signal and the j-th channel of the voice signal in the foregoing embodiment. acquisition, which will not be repeated here.

To determine the target speech signal according to the third candidate speech signal, the fourth candidate speech signal and the processed second target beam, refer to the first candidate speech signal and the second candidate speech signal in the foregoing embodiment with reference to The voice signal and the processed first target beam determine the target voice signal, which will not be repeated here.

In a possible implementation, if the interference candidate beam is not a preset beam, the target speech signal can be obtained by the following method:

The target speech signal is determined according to the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal. Specifically, the target speech signal may be determined according to the first candidate speech signal, the second candidate speech signal, and the processed first target beam with reference to the foregoing embodiment, which will not be repeated here.

In this example, de-reverberation and blind source separation are performed on the i-th third speech signal and the j-th third speech signal to obtain the first candidate speech signal and the second candidate speech signal and the noise-reduced first speech signal. The target beam is used to determine the target speech signal, which can improve the signal quality of the target speech signal.

Referring to FIG. 3 , FIG. 3 is a schematic structural diagram of a voice processing apparatus according to an embodiment of the present invention. As shown in Figure 3, the voice processing device 30 includes:

Elimination unit 301, configured to perform echo cancellation on n channels of first voice signals to obtain n channels of second voice signals, where the first voice signals are collected voice signals;

a beamforming unit 302, configured to perform beamforming on the n channels of second speech signals to obtain m channels of first beams;

an obtaining unit 303, configured to obtain interference candidate beams from the m first beams;

The processing unit 304 is configured to process the n channels of second speech signals according to the interference candidate beams to obtain a target speech signal.

In a possible implementation manner, the obtaining unit 303 is specifically configured to:

obtaining the frame-level energy corresponding to the m first beams to obtain m first frame-level energy values;

determining the average energy value of the beam according to the m first frame-level energy values;

determining a count value corresponding to the m first beams according to the m first frame-level energy values and the beam average energy value;

The beam corresponding to the largest count value among the count values corresponding to the m first beams is determined as the interference candidate beam.

In a possible implementation manner, the processing unit 304 is specifically configured to:

obtaining the smoothed energy value of the interference candidate beam;

determining a filter strength value according to the smoothed energy value;

Perform filtering processing on the n channels of second speech signals according to the filter strength value and the interference candidate beam to obtain n channels of third speech signals;

The n third voice signals are processed to obtain a target voice signal.

In a possible implementation manner, in the aspect of processing the n third voice signals to obtain the target voice signal, the processing unit 304 is specifically configured to:

Obtain the ith third voice signal and the jth third voice signal from the n third voice signals;

Perform de-reverberation and blind source separation on the i-th third speech signal and the j-th third speech signal to obtain the first candidate speech signal and the second candidate speech signal;

performing beamforming on the n channels of third speech signals to obtain m channels of second beams;

obtaining a first target beam from the m-way second beam;

performing at least noise reduction processing on the first target beam to obtain a processed first target beam;

The target speech signal is determined according to the first candidate speech signal and the second candidate speech signal and the processed first target beam.

In a possible implementation manner, the processing unit 304 is configured to:

Determine whether the interference candidate beam is a preset beam, and if the interference candidate beam is a preset beam, perform filtering processing on the n channels of second speech signals according to the interference candidate beam and the preset filtering strength to obtain n fourth voice signal;

performing beamforming on the n channels of fourth speech signals to obtain m channels of third beams;

obtaining a second target beam from the m third beams;

performing at least noise reduction processing on the second target beam to obtain a processed second target beam;

Obtain the h-th second voice signal and the k-th second voice signal from the n-way second voice signals;

Filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal to obtain a third candidate voice signal and a fourth candidate voice signal;

The target speech signal is determined according to the third candidate speech signal and the fourth candidate speech signal and the processed second target beam.

In a possible implementation manner, the processing unit 304 is further configured to:

If the interference candidate beam is not a preset beam, the target speech signal is determined according to the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal.

In this embodiment, the speech processing apparatus 300 is presented in the form of a unit. A "unit" herein may refer to an application-specific integrated circuit (ASIC), a processor and memory executing one or more software or firmware programs, an integrated logic circuit, and/or other devices that may provide the above-described functions . In addition, the above elimination unit 301 , beam forming unit 302 , acquisition unit 303 and processing unit 304 may be implemented by the processor 401 of the speech processing apparatus shown in FIG. 4 .

As shown in FIG. 4 , the voice processing apparatus 40 can be implemented with the structure shown in FIG. 4 . The voice processing apparatus 400 includes at least one processor 401 , at least one memory 402 and at least one communication interface 403 . The processor 401, the memory 402 and the communication interface 403 are connected through the communication bus and complete the communication with each other.

The processor 401 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the programs in the above solutions.

The communication interface 403 is used to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Networks (Wireless Local Area Networks, WLAN).

Memory 402 may be read-only memory (ROM) or other type of static storage device that can store static information and instructions, random access memory (RAM), or other type of static storage device that can store information and instructions The dynamic storage device can also be an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, optical disk storage ( including compact discs, laser discs, compact discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or capable of carrying or storing desired program code in the form of instructions or data structures and capable of being stored by a computer any other medium taken, but not limited to this. The memory can exist independently and be connected to the processor through a bus. The memory can also be integrated with the processor.

Wherein, the memory 402 is used for storing the application code for executing the above solution, and the execution is controlled by the processor 401 . The processor 401 is configured to execute the application code stored in the memory 402 .

The code stored in the memory 402 can execute the voice processing method provided above, and perform echo cancellation on n-channel first voice signals to obtain n-channel second voice signals, and the first voice signals are the collected voice signals; The n channels of second speech signals are beamformed to obtain m channels of first beams; interference candidate beams are obtained from the m channels of first beams; the n channels of second speech signals are processed according to the interference candidate beams, to get the target speech signal.

An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium may store a program, and when the program is executed, the program includes part or all of the steps of any of the speech processing methods described in the above method embodiments .

It should be noted that, for the sake of simple description, the foregoing method embodiments are all expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence. As in accordance with the present invention, certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative, for example, the division of the units is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components may be combined or Integration into another system, or some features can be ignored, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be in electrical or other forms.

The units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented as a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable memory. Based on such understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, Several instructions are included to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable memory, and the memory can include: a flash disk , Read-only memory (English: Read-Only Memory, referred to as: ROM), random access device (English: Random Access Memory, referred to as: RAM), magnetic disk or optical disk, etc.

The embodiments of the present invention have been introduced in detail above, and specific examples are used to illustrate the principles and implementations of the present invention. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the present invention; at the same time, for Persons of ordinary skill in the art, according to the idea of the present invention, will have changes in the specific embodiments and application scope. To sum up, the content of this description should not be construed as a limitation of the present invention.

Claims (14)

1. a speech processing method, is characterized in that, described method comprises: Performing echo cancellation on n-channel first voice signals to obtain n-channel second voice signals, where the first voice signals are collected voice signals; performing beamforming on the n channels of second speech signals to obtain m channels of first beams; obtaining interference candidate beams from the m first beams; The n channels of second speech signals are processed according to the interference candidate beams to obtain a target speech signal. 2 . The method according to claim 1 , wherein the obtaining interference candidate beams from the m-channel first beams comprises: 2 . obtaining the frame-level energy corresponding to the m first beams to obtain m first frame-level energy values; determining the average energy value of the beam according to the m first frame-level energy values; determining a count value corresponding to the m first beams according to the m first frame-level energy values and the beam average energy value; The beam corresponding to the largest count value among the count values corresponding to the m first beams is determined as the interference candidate beam. 3. The method according to claim 1 or 2, wherein the processing of the n channels of second voice signals according to the interference candidate beams to obtain a target voice signal, comprising: obtaining the smoothed energy value of the interference candidate beam; determining a filter strength value according to the smoothed energy value; Perform filtering processing on the n channels of second speech signals according to the filter strength value and the interference candidate beam to obtain n channels of third speech signals; The n third voice signals are processed to obtain a target voice signal. 4. The method according to claim 3, wherein the processing of the n third voice signals to obtain a target voice signal comprises: Obtain the ith third voice signal and the jth third voice signal from the n third voice signals; Perform de-reverberation and blind source separation on the i-th third speech signal and the j-th third speech signal to obtain the first candidate speech signal and the second candidate speech signal; performing beamforming on the n channels of third speech signals to obtain m channels of second beams; obtaining a first target beam from the m-way second beam; performing at least noise reduction processing on the first target beam to obtain a processed first target beam; The target speech signal is determined according to the first candidate speech signal and the second candidate speech signal and the processed first target beam. 5. The method according to claim 1 or 2, wherein the processing of the n channels of second speech signals according to the interference candidate beams to obtain a target speech signal, comprising: Determine whether the interference candidate beam is a preset beam, and if the interference candidate beam is a preset beam, perform filtering processing on the n channels of second speech signals according to the interference candidate beam and the preset filtering strength to obtain n fourth voice signal; performing beamforming on the n channels of fourth speech signals to obtain m channels of third beams; obtaining a second target beam from the m third beams; performing at least noise reduction processing on the second target beam to obtain a processed second target beam; Obtain the h-th second voice signal and the k-th second voice signal from the n-way second voice signals; Filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal to obtain a third candidate voice signal and a fourth candidate voice signal; The target speech signal is determined according to the third candidate speech signal and the fourth candidate speech signal and the processed second target beam. 6. The method according to claim 5, wherein the method further comprises: If the interference candidate beam is not a preset beam, the target speech signal is determined according to the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal. 7. A voice processing device, wherein the device comprises: an elimination unit, configured to perform echo cancellation on the n-channel first voice signals to obtain n-channel second voice signals, where the first voice signals are the collected voice signals; a beamforming unit, configured to perform beamforming on the n channels of second speech signals to obtain m channels of first beams; an obtaining unit, configured to obtain interference candidate beams from the m first beams; A processing unit, configured to process the n channels of second speech signals according to the interference candidate beams to obtain a target speech signal. 8. The device according to claim 7, wherein the acquiring unit is specifically configured to: obtaining the frame-level energy corresponding to the m first beams to obtain m first frame-level energy values; determining an average beam energy value according to the m first frame-level energy values; determining a count value corresponding to the m first beams according to the m first frame-level energy values and the beam average energy value; The beam corresponding to the largest count value among the count values corresponding to the m first beams is determined as the interference candidate beam. 9. The device according to claim 7 or 8, wherein the processing unit is specifically configured to: obtaining the smoothed energy value of the interference candidate beam; determining a filter strength value according to the smoothed energy value; Perform filtering processing on the n channels of second speech signals according to the filter strength value and the interference candidate beam to obtain n channels of third speech signals; The n third voice signals are processed to obtain a target voice signal. 10. The device according to claim 9, wherein, in the aspect of processing the n third voice signals to obtain the target voice signal, the processing unit is specifically used for: Obtain the ith third voice signal and the jth third voice signal from the n third voice signals; Perform de-reverberation and blind source separation on the i-th third voice signal and the j-th third voice signal to obtain the first candidate voice signal and the second candidate voice signal; performing beamforming on the n channels of third speech signals to obtain m channels of second beams; obtaining a first target beam from the m-way second beam; performing at least noise reduction processing on the first target beam to obtain a processed first target beam; The target speech signal is determined according to the first candidate speech signal and the second candidate speech signal and the processed first target beam. 11. The apparatus according to claim 7 or 8, wherein the processing unit is configured to: Determine whether the interference candidate beam is a preset beam, and if the interference candidate beam is a preset beam, perform filtering processing on the n channels of second speech signals according to the interference candidate beam and the preset filtering strength to obtain n fourth voice signal; performing beamforming on the n channels of fourth speech signals to obtain m channels of third beams; obtaining a second target beam from the m third beams; performing at least noise reduction processing on the second target beam to obtain a processed second target beam; Obtain the h-th second voice signal and the k-th second voice signal from the n-way second voice signals; Filtering and blind source separation are performed on the h-th second voice signal and the k-th second voice signal to obtain a third candidate voice signal and a fourth candidate voice signal; The target speech signal is determined according to the third candidate speech signal and the fourth candidate speech signal and the processed second target beam. 12. The apparatus according to claim 11, wherein the processing unit is further configured to: If the interference candidate beam is not a preset beam, the target speech signal is determined according to the interference candidate beam, the third candidate speech signal and the fourth candidate speech signal. 13. A voice processing device, comprising: memory for storing instructions; and at least one processor coupled to the memory; Wherein, when the at least one processor executes the instructions, the instructions cause the processor to perform the method of any one of claims 1-6. 14. A computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, the computer program comprising program instructions that, when executed by a processor, cause the processor to execute The method of any one of claims 1-6.

Images