Voice noise reduction method and system, electronic equipment and product

Abstract

The invention discloses a voice noise reduction method and system, electronic equipment and a product. The method comprises the following steps: acquiring an array voice signal and obtaining a first target pickup area corresponding to a first target sound source; suppressing voice signals outside the first target pickup area in the array voice signals based on the first guiding matrix to obtain a first target voice signal containing a first residual noise signal; voice signals corresponding to a first target pickup area in the array voice signals are removed based on the first blocking matrix, and first noise signals are obtained; screening out a first adaptive matrix from a plurality of adaptive matrixes based on the first target pickup area, wherein the first adaptive matrix is used for converting the first noise signal into a noise signal which is the same as the first residual noise signal; and removing a first residual noise signal in the first target voice signal based on the first noise signal subjected to the first adaptive matrix conversion to obtain a first noise-reduced voice signal. And keeping the voice noise reduction level unchanged when the sound source direction of the first target sound source is changed.

Description

Technical Field

[0001] The present application relates to the field of speech processing technology, and in particular to a speech noise reduction method, system, electronic equipment and product. Background Art

[0002] Due to the rapid development of network and audio technology, people are no longer limited to face-to-face meetings. They can also participate in meetings remotely through online meetings.

[0003] In order to improve the voice quality of online meetings, many voice noise reduction methods for conference scenarios have been studied, such as using directional microphones to suppress noise other than the speaker. However, this method requires the deployment of more microphones, which is costly and inflexible. At present, the commonly used method is to use an adaptive noise reduction algorithm based on spatial domain features to collect speakers in a specific direction, suppress noise sources in other directions, and increase the signal-to-noise ratio of the conference voice. However, this method has an unstable noise reduction level when the direction of the sound source changes, which seriously reduces the voice quality of online meetings. Summary of the invention

[0004] The present application provides a speech noise reduction method, system, electronic device and product to solve the problem of unstable noise reduction level of existing speech noise reduction methods; an adapted first adaptive matrix is ​​selected from a plurality of pre-set adaptive matrices through a first target pickup area of ​​a first target sound source for noise reduction processing, and when the direction of the sound source changes, the speech noise reduction level can be kept from decreasing, thereby improving the stability of the speech noise reduction level and improving the speech output quality.

[0005] In a first aspect, an embodiment of the present application provides a method for reducing speech noise, including:

[0006] Acquire the array voice signals of different sound pickup areas collected by the microphone array and obtain a first target sound pickup area corresponding to the first target sound source;

[0007] Suppressing the speech signal outside the first target sound pickup area in the array speech signal based on the first steering matrix to obtain a first target speech signal containing a first residual noise signal;

[0008] Removing the speech signal corresponding to the first target sound pickup area from the array speech signal based on the first blocking matrix to obtain a first noise signal;

[0009] Screening out a first adaptive matrix from a plurality of preset adaptive matrices based on the first target sound pickup area, and converting the first noise signal into a noise signal identical to the first residual noise signal based on the first adaptive matrix;

[0010] The first residual noise signal in the first target speech signal is removed based on the first noise signal converted by the first adaptive matrix to obtain a first noise reduction speech signal corresponding to the first target sound source.

[0011] In a second aspect, an embodiment of the present application provides a speech noise reduction system, including: a first steering matrix, a first blocking matrix, a plurality of adaptive matrices and a first noise reduction module; wherein,

[0012] The first steering matrix is ​​used to suppress the speech signal outside the first target sound pickup area corresponding to the first target sound source in the array speech signal to obtain a first target speech signal containing a first residual noise signal; the array speech signal is obtained by collecting speech signals in different sound pickup areas by the microphone array;

[0013] The first blocking matrix is ​​used to suppress the voice signal corresponding to the first target sound pickup area in the array voice signal to obtain a first noise signal;

[0014] A first adaptive matrix is ​​used to convert the first noise signal into a noise signal identical to the first residual noise signal; the first adaptive matrix is ​​obtained by screening a plurality of pre-set adaptive matrices based on the first target sound pickup area;

[0015] The first noise reduction module is used to remove the first residual noise signal in the first target speech signal based on the first noise signal converted by the first adaptive matrix, so as to obtain a first noise-reduced speech signal corresponding to the first target sound source.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, the electronic device comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the speech noise reduction method described in any embodiment of the present application.

[0020] In a fourth aspect, an embodiment of the present application provides a computer program product including a computer program, which, when executed by a processor, implements the speech noise reduction method described in any embodiment of the present application.

[0021] The technical solution of the embodiment of the present application is to obtain the array voice signals of different pickup areas collected by the microphone array and obtain the first target pickup area corresponding to the first target sound source; based on the first guide matrix, the voice signals outside the first target pickup area in the array voice signal are suppressed to obtain the first target voice signal containing the first residual noise signal; based on the first blocking matrix, the voice signals corresponding to the first target pickup area in the array voice signal are removed to obtain the first noise signal; based on the first target pickup area, the first adaptive matrix is ​​selected from the pre-set multiple adaptive matrices, and the first noise signal is converted into the same noise signal as the first residual noise signal based on the first adaptive matrix; based on the first noise signal converted by the first adaptive matrix, the first residual noise signal in the first target voice signal is removed to obtain the first noise reduction voice signal corresponding to the first target sound source. The first adaptive matrix is ​​selected from the pre-set multiple adaptive matrices for noise reduction processing through the first target pickup area of ​​the first target sound source; when the direction of the sound source changes, the first adaptive matrix used is switched according to the change of the first target pickup area, so that the voice noise reduction level can be kept unchanged, which solves the problem of unstable noise reduction level of the existing voice noise reduction method, improves the stability of the voice noise reduction level, and improves the voice output quality.

[0022] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A schematic diagram of the principle of a generalized sidelobe cancellation algorithm provided in the prior art;

[0025] Figure 2 A flow chart of a speech noise reduction method provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of the principles of a speech noise reduction method and an adaptive matrix update method in a single sound source mode provided in an embodiment of the present application;

[0027] Figure 4 A schematic diagram of the principle of a speech noise reduction method in a multi-sound source mode provided in an embodiment of the present application;

[0028] Figure 5 A schematic diagram of the principle of a method for combining speech noise reduction and adaptive matrix updating in a multi-sound source mode provided in an embodiment of the present application;

[0029] Figure 6 A schematic diagram of the structure of a speech noise reduction system provided in an embodiment of the present application;

[0030] Figure 7 It is a structural schematic diagram of an electronic device for implementing the speech noise reduction method of an embodiment of the present application. DETAILED DESCRIPTION

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] Figure 1 The following is a schematic diagram of the principle of a generalized sidelobe cancellation algorithm provided in the prior art. The generalized sidelobe cancellation algorithm is an algorithm that improves signal quality by eliminating sidelobe noise. Figure 1As shown, the principle of the generalized sidelobe cancellation algorithm is: the coherent noise signal of the array voice signal collected by the microphone array outside the target pickup area is suppressed by the steering matrix, but because the coherent noise signal outside the target pickup area cannot be completely suppressed, the steering matrix output signal contains the voice signal and the residual noise signal in the target pickup area. The voice signal in the target pickup area of ​​the array voice signal is suppressed by the blocking matrix to obtain the noise signal; and the noise signal is converted into a signal identical to the residual noise signal by the adaptive matrix; based on the noise signal converted by the adaptive matrix, the residual noise signal in the steering matrix output signal is removed to obtain the denoised voice signal of the array voice signal in the target pickup area, thereby achieving denoising of the array voice signal. In addition, the denoised voice signal without voice can also be used to update the weight of the adaptive matrix so that the adaptive matrix is ​​adaptively matched with the voice environment.

[0034] However, the existing generalized sidelobe cancellation algorithm has the following problems:

[0035] First, when the direction of the speech signal changes, the adaptive matrix no longer adapts to the speech signal, resulting in a decrease in the noise suppression ability of the generalized sidelobe cancellation algorithm. It is necessary to wait for the adaptive matrix to complete the weight update before the noise suppression ability can be restored, resulting in unstable noise reduction quality.

[0036] Second, most generalized sidelobe cancellation algorithms use speech signals during speechless periods to update the weights of the adaptive matrix. The update time depends on the length of time when no one is speaking, resulting in slow update speed and unstable update effect. Especially for long-term speech, there may be a risk of long-term non-update or invalid update, which seriously affects the noise reduction quality of conference speech.

[0037] Third, in a conference scenario, there may be multiple speakers speaking at the same time. The existing generalized sidelobe cancellation algorithm needs to use a single beam to switch back and forth between multiple sound sources in the multi-sound source mode, which may cause the sound of the sound source to fluctuate or even lose some voice information.

[0038] In response to the above problems, an embodiment of the present application provides a method for speech noise reduction, by setting multiple adaptive matrices based on the number of pickup areas of a microphone array. When the pickup area corresponding to the speech signal changes, the adaptive matrix corresponding to the pickup area can be selected to convert the noise, thereby improving the stability of speech noise reduction. And while someone is speaking, the adaptive matrix is ​​updated with the collected speech signal to reduce the time for updating the adaptive matrix, and can maintain good speech noise reduction quality for long-term speeches. In addition, for scenarios where multiple speakers are speaking at the same time, it is also possible to effectively reduce the noise of speech from multiple sound sources, keep the speakers' voices normal, and reduce the loss of speech information.

[0039] Next, a speech noise reduction method provided by the present application is described in detail in conjunction with a specific implementation method.

[0040] Figure 2 This is a flow chart of a method for reducing the noise of speech provided in an embodiment of the present application. This embodiment can be applied to the case of reducing the noise of speech collected by a microphone in a voice conference scenario. The method can be performed by a speech noise reduction system. The speech noise reduction system can be implemented in the form of hardware and / or software. The speech noise reduction system can be configured in an electronic device. Figure 2 As shown, the method includes:

[0041] S110, acquiring array voice signals collected by a microphone array in different sound pickup areas and obtaining a first target sound pickup area corresponding to a first target sound source.

[0042] Among them, the microphone array can be considered as an array formed by a group of microphones located at different positions in space and arranged in a certain shape, which is used for spatial sampling of spatially propagated sound signals. The pickup area can be considered as a non-overlapping area divided according to the coverage range of the pickup beam formed by the microphone array through beamforming technology. Generally, a microphone array has multiple pickup areas. The array voice signal can be considered as a mixed voice signal collected from different pickup areas of the microphone array. The array voice signal may include: a first target sound source signal and a noise signal. The first target sound source can be understood as the source of the sound. In a video conference, the first target sound source can be considered as the conference speaker. The first target pickup area can be understood as the pickup area where the pickup beam covers the location of the first target sound source.

[0043] Specifically, array voice signals of different sound pickup areas collected by the microphone array are obtained, and the array voice signals are calculated using a sound source localization algorithm to obtain a first target sound pickup area corresponding to a first target sound source.

[0044] Exemplarily, the sound source localization algorithm may adopt a phase-transformed weighted directional response power (SRP-PHAT) algorithm and a multiple signal classification (MUSIC) algorithm, etc., which is not limited in the embodiments of the present application.

[0045] S120: Suppress the speech signal outside the first target sound pickup area in the array speech signal based on the first steering matrix to obtain a first target speech signal containing a first residual noise signal.

[0046] Among them, the first target speech signal can be considered as the speech signal output by the first steering matrix. The first residual noise signal can be considered as a partial coherent noise signal that cannot be removed by the first steering matrix. The first steering matrix can be considered as the steering matrix corresponding to the first target pickup area. The function of the first steering matrix is ​​to form a fixed beam in the direction of the first target pickup area, and use the fixed beam to retain the signal in the first target pickup area as much as possible, while suppressing the coherent noise signal outside the first target pickup area. Exemplarily, the calculation of the first steering matrix can be designed using a constant-width beam design method, the beam width is equal to the width of the first target pickup area, and the fixed beam is distortion-free in the first target pickup area. The constant-width beam design method can, for example, be a broadband beamforming method and a spatial resampling method based on focusing transformation, which is not limited or elaborated in the embodiments of the present application.

[0047] Specifically, the fixed beam formed by the first steering matrix in the direction of the first target sound pickup area corresponding to the first target sound source suppresses the speech signal outside the first target sound pickup area. However, due to the limitation of the beam design, the fixed beam cannot completely suppress the speech signal outside the first target sound pickup area. The speech signal remaining in other sound pickup areas outside the first target sound pickup area can be considered as the first residual noise signal relative to the speech signal in the first target sound pickup area. Therefore, the first target speech signal output by the first steering matrix contains the first residual noise signal.

[0048] S130. Remove the speech signal corresponding to the first target sound pickup area in the array speech signal based on the first blocking matrix to obtain a first noise signal.

[0049] Among them, the first noise signal can be considered as the signal output by the first blocking matrix, and can further be considered as the signal of other pickup areas outside the first target pickup area output by the first blocking matrix. The first blocking matrix can be considered as a blocking matrix corresponding to the first target pickup area, and its function is to remove the voice signal of the first target pickup area in the array voice signal as much as possible, and only retain the coherent noise signal outside the first target pickup area. Exemplarily, the first blocking matrix is ​​generated by a second-order cone optimization design, and is complementary to the constant-width beam formed by the first steering matrix in terms of beam shape. The embodiment of the present application does not limit or elaborate on the design method of the first blocking matrix.

[0050] Specifically, the array voice signal is processed based on the first blocking matrix to form a gain depression in the first target sound pickup area, the voice signal corresponding to the first target sound pickup area in the array voice signal is removed, and the noise signal outside the first target sound pickup area is retained as the first noise signal.

[0051] S140: Screen out a first adaptive matrix from a plurality of pre-set adaptive matrices based on the first target sound pickup area, and convert the first noise signal into a noise signal identical to the first residual noise signal based on the first adaptive matrix.

[0052] In this embodiment, a plurality of adaptive matrices are pre-set. Optionally, the number of adaptive matrices can be determined based on the number of pickup areas of the microphone array. The first adaptive matrix can be considered as an adaptive matrix corresponding to the first target pickup area. The function of the first adaptive matrix is ​​to convert the first noise signal output by the first blocking matrix into a noise signal identical to the first residual noise signal that cannot be removed by the first guiding matrix.

[0053] Specifically, based on the first target sound pickup area corresponding to the first target sound source, a first adaptive matrix corresponding to the first target sound pickup area is selected from a plurality of pre-set adaptive matrices. Based on the first adaptive matrix, the first noise signal output by the first blocking matrix is ​​converted so that the first noise signal is converted into a noise signal identical to the first residual noise signal contained in the first target speech signal output by the first steering matrix.

[0054] S150: Remove a first residual noise signal in a first target speech signal based on the first noise signal converted by a first adaptive matrix to obtain a first noise-reduced speech signal corresponding to a first target sound source.

[0055] The first noise reduction speech signal may be considered as a speech signal obtained by removing noise from the first target speech signal.

[0056] Specifically, since the first noise signal converted by the first adaptive matrix is ​​the same as the first residual noise signal in the first target voice signal, the first noise-reduced voice signal can be obtained by removing the first residual noise signal in the first target voice signal from the first noise signal converted by the first adaptive matrix. The first noise-reduced voice signal removes the voice signal outside the first target sound pickup area relative to the array voice signal, and only retains the voice signal in the first target sound pickup area.

[0057] The technical solution of the embodiment of the present application is to obtain the array voice signals of different pickup areas collected by the microphone array and obtain the first target pickup area corresponding to the first target sound source; based on the first guide matrix, the voice signals outside the first target pickup area in the array voice signal are suppressed to obtain the first target voice signal containing the first residual noise signal; based on the first blocking matrix, the voice signals corresponding to the first target pickup area in the array voice signal are removed to obtain the first noise signal; based on the first target pickup area, the first adaptive matrix is ​​selected from the pre-set multiple adaptive matrices, and the first noise signal is converted into the same noise signal as the first residual noise signal based on the first adaptive matrix; based on the first noise signal converted by the first adaptive matrix, the first residual noise signal in the first target voice signal is removed to obtain the first noise reduction voice signal corresponding to the first target sound source. The first adaptive matrix is ​​selected from the pre-set multiple adaptive matrices for noise reduction processing through the first target pickup area of ​​the first target sound source; when the direction of the sound source changes, the first adaptive matrix used is switched according to the change of the first target pickup area, so that the voice noise reduction level can be kept unchanged, which solves the problem of unstable noise reduction level of the existing voice noise reduction method, improves the stability of the voice noise reduction level, and improves the voice output quality.

[0058] As an optional embodiment of the embodiment of the present application, the step of removing the first residual noise signal in the first target speech signal based on the first noise signal converted by the first adaptive matrix to obtain a first noise-reduced speech signal corresponding to the first target sound source includes:

[0059] S151. Use a first multiplier with a gain coefficient of -1 to transform the first noise signal transformed by the first adaptive matrix to obtain a first noise inverse signal.

[0060] The first anti-noise signal may be understood as an anti-signal of the first noise signal converted by the first adaptive matrix, and the first anti-noise signal and the first noise signal converted by the first adaptive matrix may be added to cancel each other out.

[0061] Specifically, a first multiplier with a gain coefficient of -1 is used to transform the first noise signal converted by the first adaptive matrix to obtain a first anti-noise signal that can offset the first noise signal converted by the first adaptive matrix.

[0062] S152: Use an adder to add the first anti-noise signal and the first target voice signal to obtain a first noise reduction voice signal corresponding to the first target sound pickup area.

[0063] Specifically, since the first noise signal converted by the first adaptive matrix is ​​the same as the first residual noise signal, the first noise inverse signal and the first residual noise signal can also cancel each other out when added. An adder is used to add the first noise inverse signal and the first target voice signal containing the first residual noise signal, and the first residual noise signal in the first target voice signal is removed to obtain a first noise reduction voice signal corresponding to the first target pickup area.

[0064] This embodiment can effectively remove the first residual noise signal in the first target speech signal through a multiplier and an adder, has a simple and effective structure, and has high noise removal efficiency.

[0065] As an optional implementation of the embodiment of the present application, a second steering matrix and a second blocking matrix corresponding to the second target pickup area are also introduced into the speech noise reduction method. The speech noise reduction method also includes:

[0066] S210: Use each sound pickup area except the first target sound pickup area as a second target sound pickup area.

[0067] The second target sound pickup area may be considered as each sound pickup area not covered by the beam of the first target sound source, that is, each sound pickup area except the first target sound pickup area.

[0068] It can be understood that if the microphone array has N sound pickup areas, the second target sound pickup area is each of the N-1 sound pickup areas except the first target sound pickup area.

[0069] S220: Based on a second steering matrix corresponding to the second target sound pickup area, suppress the speech signal outside the second target sound pickup area in the array speech signal to obtain a second target speech signal containing a second residual noise signal.

[0070] Among them, the second target voice signal can be considered as the voice signal output by the second steering matrix. The second residual noise signal can be considered as a partial coherent noise signal that cannot be removed by the second steering matrix. The second steering matrix can be considered as the steering matrix corresponding to the second target pickup area. The function of the second steering matrix is ​​to form a fixed beam in the direction of the second target pickup area, and use the fixed beam to retain the signal in the second target pickup area as much as possible, while suppressing the coherent noise signal outside the second target pickup area. The embodiment of the present application also does not limit or elaborate on the design method of the second steering matrix.

[0071] Specifically, the fixed beam formed by the second steering matrix in the direction of the second target sound pickup area corresponding to the second target sound source suppresses the speech signal outside the second target sound pickup area. However, due to the limitation of the beam design, the fixed beam cannot completely suppress the speech signal outside the second target sound pickup area. The speech signal remaining in other sound pickup areas outside the second target sound pickup area can be considered as the second residual noise signal relative to the speech signal in the second target sound pickup area. Therefore, the second target speech signal output by the second steering matrix contains the second residual noise signal.

[0072] S230 , suppressing a speech signal corresponding to the second target sound pickup area in the array speech signal based on a second blocking matrix corresponding to the second target sound pickup area to obtain a second noise signal.

[0073] Among them, the second noise signal can be considered as the signal output by the second blocking matrix, and can further be considered as the signal of other pickup areas outside the second target pickup area output by the second blocking matrix. The second blocking matrix can be considered as a blocking matrix corresponding to the second target pickup area, and its function is to remove the voice signal of the second target pickup area in the array voice signal as much as possible, and only retain the coherent noise signal outside the second target pickup area. The embodiment of the present application also does not limit or elaborate on the design method of the second blocking matrix.

[0074] Specifically, the array voice signal is processed based on the second blocking matrix to form a gain depression in the second target sound pickup area, the voice signal corresponding to the second target sound pickup area in the array voice signal is removed, and the noise signal outside the second target sound pickup area is retained as the second noise signal.

[0075] S240: Screen out a second adaptive matrix from a plurality of preset adaptive matrices based on the second target sound pickup area, and convert the second noise signal into a noise signal identical to the second residual noise signal based on the second adaptive matrix.

[0076] The second adaptive matrix can be considered as an adaptive matrix corresponding to the second target sound pickup area. The second adaptive matrix is ​​used to convert the second noise signal output by the second blocking matrix into a noise signal identical to the second residual noise signal that cannot be removed by the second steering matrix.

[0077] Specifically, based on the second target sound pickup area corresponding to the second target sound source, a second adaptive matrix corresponding to the second target sound pickup area is selected from a plurality of pre-set adaptive matrices. Based on the second adaptive matrix, the second noise signal output by the second blocking matrix is ​​converted so that the second noise signal is converted into a noise signal identical to the second residual noise signal contained in the second target speech signal output by the second steering matrix.

[0078] S250: Remove a second residual noise signal from a second target speech signal based on the second noise signal converted by a second adaptive matrix to obtain a second noise-reduced speech signal corresponding to a second target sound pickup area.

[0079] The second noise reduction speech signal may be considered as a speech signal obtained after noise is removed from the second target speech signal.

[0080] Specifically, since the first noise signal converted by the second adaptive matrix is ​​the same as the second residual noise signal in the second target voice signal, the first noise reduction voice signal can be obtained by removing the second residual noise signal in the second target voice signal from the second noise signal converted by the second adaptive matrix. The second noise reduction voice signal removes the voice signal outside the second target pickup area relative to the array voice signal, and only retains the voice signal in the second target pickup area.

[0081] As an optional implementation of the embodiment of the present application, the step of removing the second residual noise signal in the second target voice signal based on the second noise signal converted by the second adaptive matrix to obtain a second noise reduction voice signal corresponding to the second target pickup area includes:

[0082] S251. Use a second multiplier with a gain coefficient of -1 to transform the second noise signal transformed by the second adaptive matrix to obtain a second noise inverse signal.

[0083] The second anti-noise signal can be understood as the anti-signal of the second noise signal converted by the second adaptive matrix, and the second anti-noise signal and the second noise signal converted by the second adaptive matrix can cancel each other out when added together.

[0084] Specifically, a second multiplier with a gain coefficient of -1 is used to transform the second noise signal converted by the second adaptive matrix to obtain a second anti-noise signal that can offset the second noise signal converted by the second adaptive matrix.

[0085] S252: Use an adder to add the second anti-noise signal and the second target voice signal to obtain a second noise reduction voice signal corresponding to the second target sound pickup area.

[0086] Specifically, since the second noise signal converted by the second adaptive matrix is ​​the same as the second residual noise signal, the second noise inverse signal and the second residual noise signal can also cancel each other out when added. An adder is used to add the second noise inverse signal and the second target voice signal including the second residual noise signal, and the second residual noise signal in the second target voice signal is removed to obtain a second noise reduction voice signal corresponding to the second target pickup area.

[0087] This embodiment can effectively remove the second residual noise signal in the second target speech signal through a multiplier and an adder, has a simple and effective structure, and has high noise removal efficiency.

[0088] As an optional implementation of the embodiment of the present application, the number of the adaptive matrices, the total number of the first steering matrices and the second steering matrices, and the total number of the first blocking matrices and the second blocking matrices are the same as the number of pickup areas of the microphone array. The first target pickup area corresponds to a first steering matrix, a first blocking sentence matrix, and a first adaptive matrix; each second target pickup area corresponds to a second steering matrix, a second blocking sentence matrix, and a second adaptive matrix.

[0089] In this embodiment, the second noise reduction speech signal has two functions. The first function is to use the second noise reduction speech signal to update the second adaptive matrix corresponding to the second target sound pickup area when the speech signal corresponding to the second target sound pickup area is a non-sound source signal. The second function is to use it to mix with the noise reduction speech of other target sound pickup areas as the noise reduction result output by the speech noise reduction method when the speech signal corresponding to the second target sound pickup area is a sound source signal.

[0090] As an optional implementation of the embodiment of the present application, a process for updating an adaptive matrix in a speech noise reduction method is also provided. The specific implementation process of each second adaptive matrix update also includes:

[0091] S260: If the speech signal corresponding to the second target sound pickup area is a non-sound source signal, update the weights of the second adaptive matrix based on the second noise reduction speech signal and the second noise signal.

[0092] The non-sound source signal may be understood as a signal that is not a sound source, and may include, for example, device noise and environmental noise.

[0093] Specifically, if the voice signal corresponding to the second target sound pickup area is a non-sound source signal, the second noise reduction voice signal is used to update the weight of the second adaptive matrix. Based on the second noise reduction voice signal corresponding to the second target sound pickup area and the second blocking matrix corresponding to the second target sound pickup area outputs a second noise signal, the weight of the second adaptive matrix corresponding to the second target sound pickup area is updated.

[0094] It can be understood that if the voice signal corresponding to the second target sound pickup area switches from a non-sound source signal to a sound source signal, the second noise reduction voice signal corresponding to the second target sound pickup area is no longer used to mix with the second noise reduction voice signal to obtain the target noise reduction voice signal; instead, it is converted to be used to update the weights of the second adaptive matrix with the second noise reduction voice signal.

[0095] Exemplarily, the method for determining whether the speech signal corresponding to the second target sound pickup area is a non-sound source signal or a sound source signal may be: for each second target sound pickup area, the array speech signal is calculated using a sound source localization algorithm to determine whether there is a target sound source in the second target sound pickup area. If there is no target sound source in the second target sound pickup area, the speech signal corresponding to the second target sound pickup area is considered to be a non-sound source signal. If there is a target sound source in the second target sound pickup area, the speech signal corresponding to the second target sound pickup area is considered to be a sound source signal.

[0096] Exemplarily, the calculation method for updating the weights of the second adaptive matrix based on the second noise reduction speech signal and the second noise signal can be specifically expressed as: G=G+conj(B)×Y×μ; wherein G represents the weights of the adaptive matrix, B is the second noise signal output by the second blocking matrix, Y is the second noise reduction speech signal, and conj is the conjugate operator.

[0097] The main idea of ​​the embodiment of the present application is to use the array voice signal to update the weights of the adaptive matrix corresponding to the second target pickup area in the non-sound source direction, the second noise reduction voice signal and the second noise signal corresponding to the second target pickup area in the non-sound source direction, while performing noise reduction processing on the sound source signal, so that the adaptive matrix can be updated in time based on the array voice signal collected without a sound source or with a sound source, without waiting for the array voice signal collected without a sound source to be updated, and is not limited by whether there is a sound source in the array voice signal and the duration of the sound source, thereby improving the update efficiency of the adaptive matrix. When the direction of the sound source changes to the second target pickup area, there is no need to wait for a long time for the second adaptive matrix to complete the weight update, thereby improving the fluency of the voice output. And for the scene where the audio signal is collected for a long time, the adaptive matrix can also be updated in time, thereby effectively improving the stability of voice noise reduction.

[0098] In an optional embodiment, the updating of the weights of the second adaptive matrix based on the second noise reduction speech signal and the second noise signal is performed in parallel or in series.

[0099] In this embodiment, the first second adaptive matrix corresponding to each second target sound pickup area can update the weights of the second adaptive matrix corresponding to each second target sound pickup area in a parallel or serial manner during the noise reduction processing of the first target sound source corresponding to the first target sound pickup area in the array voice signal.

[0100] It should be noted that when the weights of the second adaptive matrix are updated in parallel, the array speech signals at the same time are used; and when the weights of the second adaptive matrix are updated in serial, the array speech signals at different times are used.

[0101] The weights of the second adaptive matrices corresponding to each second target pickup area are updated in parallel, which can further improve the update efficiency of each second adaptive matrix, thereby further reducing the noise reduction waiting time and improving the stability of the noise reduction level when the sound source direction changes. In some video conferencing scenarios, the sound source direction does not change frequently. The weights of the second adaptive matrices corresponding to each second target pickup area are updated in serial mode. Under the condition of ensuring a certain update efficiency, the computing resources occupied by the adaptive matrix update process can be reduced.

[0102] As an optional implementation of the embodiment of the present application, a method for reducing speech noise for multiple sound sources is provided. The specific implementation process of the method also includes:

[0103] S270: If the voice signal corresponding to the second target sound pickup area is a sound source signal, mix the first noise reduction voice signal and the second noise reduction voice signal to obtain a target noise reduction voice signal of the array voice signal.

[0104] The sound source signal can be understood as a signal serving as a sound source. The target noise reduction speech signal can be understood as a mixed signal of noise reduction speech signals of multiple sound source signals in the array speech signal.

[0105] Specifically, if the voice signal corresponding to the second target sound pickup area is a sound source signal, the first noise reduction voice signal is used to output the noise reduction voice. The first noise reduction voice signal corresponding to the voice signal of the first target sound pickup area in the array voice signal and the second noise reduction voice signal corresponding to the voice signal of the second target sound pickup area are mixed to obtain the target noise reduction voice signal.

[0106] It can be understood that if the voice signal corresponding to the second target pickup area switches from a sound source signal to a non-sound source signal, the second adaptive matrix corresponding to the second target pickup area will stop updating, and the second noise reduction voice signal corresponding to the second target pickup area will no longer be used to update the second adaptive matrix, but will be converted into and mixed with the second noise reduction voice signal to obtain the target noise reduction voice signal.

[0107] Exemplarily, the first noise reduction voice signal and the second noise reduction voice signal can be mixed by adding the first noise reduction voice signal and the second noise reduction voice signal using an adder; or the first noise reduction voice signal and the second noise reduction voice signal can be placed in sequence on different audio tracks, and the volume and balance of each audio track can be adjusted; and the mixing and balance of different voice signals can be achieved by adjusting parameters such as the superposition method and volume level of each voice signal on the mixing track.

[0108] The embodiment of the present application performs noise reduction processing on the sound source signal corresponding to each second target pickup area in the array voice signal to obtain a second noise-reduced voice signal, and mixes the first noise-reduced voice signal and the second noise-reduced voice signal, so as to effectively perform noise reduction processing on each sound source signal in the array voice signal, overcome the problem that a beam switches back and forth between multiple sound sources, causing the voice sound to be unstable and voice information to be easily lost, and improve the stability of voice noise reduction in the multi-sound source mode.

[0109] In an optional embodiment, if the voice signals corresponding to each second target sound pickup area are all non-sound source signals, that is, the array voice signal contains only one first target sound source and no other sound sources. In application scenarios such as video conferencing, this mode containing only one sound source in the same sound sampling period can be considered as a single sound source mode. Figure 3 The present invention provides a method for reducing the noise of a speech signal in a single-source mode and a schematic diagram of the principle of an adaptive matrix update method. Figure 3 The figure only illustrates the updating process of the second adaptive matrix corresponding to one second target sound pickup area. It can be understood that there may be multiple second target sound pickup areas in the embodiment of the present application, and the updating principle of the second adaptive matrix corresponding to each second target sound pickup area is the same. Figure 3 As shown, the main idea of ​​the speech noise reduction method in the single sound source mode is: using the first guide matrix, the first blocking matrix and the first adaptive matrix corresponding to the first target pickup area where the sound source signal is located, the array speech signal is subjected to noise reduction processing to obtain the first noise-reduced speech signal. By selecting the first adaptive matrix from the pre-set multiple adaptive matrices according to the first target pickup area where the first target sound source is located, the speech noise reduction level can be maintained without decreasing when the direction of the sound source changes, thereby improving the stability of the speech noise reduction level and improving the speech output quality.

[0110] The main idea of ​​the method for updating the adaptive matrix is: for each adaptive matrix corresponding to the second target sound pickup area outside the first target sound pickup area corresponding to the first target sound source, the weight of the adaptive matrix is ​​updated using the second noise reduction speech signal and the second noise signal corresponding to the second target sound pickup area of ​​the array speech signal. When the direction of the sound source changes to any second target sound pickup area, there is no need to wait for a long time for the second adaptive matrix to complete the weight update, and the speech noise reduction can be completed quickly, further improving the fluency of the speech output.

[0111] In another optional embodiment, if there is at least one second target sound pickup area corresponding to a voice signal that is a sound source signal, that is, the array voice signal contains not only a first target sound source, but also at least one second target sound source. In application scenarios such as video conferencing, this mode containing at least two sound sources in the same sound sampling period can be considered a multi-sound source mode.

[0112] Figure 4 A schematic diagram of the principle of a speech noise reduction method in a multi-sound source mode provided in an embodiment of the present application. Figure 4 The figure only illustrates the process of performing noise reduction processing on the speech signal corresponding to the second target sound pickup area when there is a sound source signal in the second target sound pickup area; it can be understood that in the embodiment of the present application, there can be at least one second target sound pickup area, and the principle of the updating process of the second adaptive matrix corresponding to each second target sound pickup area is the same. Figure 4 As shown, in the multi-source mode, the main idea of ​​the noise reduction method for the array voice signal is: for each sound source signal of the array voice signal, based on the guiding matrix corresponding to the target pickup area where the sound source signal is located, the voice signal outside the target pickup area in the array voice signal is suppressed to obtain the target voice signal containing the residual noise signal; based on the blocking matrix of the target pickup area where the sound source signal is located, the voice signal corresponding to the target pickup area in the array voice signal is removed to obtain the noise signal; based on the target pickup area where the sound source signal is located, the target adaptive matrix is ​​selected from the pre-set multiple adaptive matrices, and the noise signal is converted into a noise signal identical to the residual noise signal based on the target adaptive matrix; based on the noise signal converted by the target adaptive matrix, the residual noise signal in the target voice signal is removed to obtain the noise reduction voice signal corresponding to the target sound source, and the noise reduction voice signal corresponding to the target sound source is mixed to obtain the target noise reduction voice signal. It can effectively perform noise reduction processing on the multi-source signals of the array voice signal respectively, overcome the problem that a beam switches back and forth between multiple sound sources, causing the voice sound to be unstable and easy to lose voice information, improve the stability of voice noise reduction in the multi-source mode, and improve the voice output quality.

[0113] Figure 5A schematic diagram of the principle of a method for combining speech noise reduction and adaptive matrix update in a multi-source mode provided in an embodiment of the present application. Figure 5 The figure only illustrates the process of noise reduction processing for the speech signal corresponding to the second target pickup area when there is a sound source signal in the second target pickup area; and the process of updating the second adaptive matrix corresponding to the second target pickup area when there is no sound source signal in the second target pickup area. It can be understood that in the multi-sound source mode, there can be at least one second target pickup area with a sound source signal, and the principle of noise reduction processing for the speech signal corresponding to each second target pickup area is the same. Figure 5 As shown, in the multi-sound source mode, the main idea of ​​the method for updating the adaptive matrix is: while performing speech denoising on multiple sound source signals, the second denoised speech signal and the second noise signal corresponding to the second target pickup area that does not contain the sound source signal are used to timely update the weights of the adaptive matrix that is not involved in the speech denoising process of the sound source signal. Therefore, when the direction of any sound source changes to other second target pickup areas, there is no need to wait for a long time for the weights of the second adaptive matrix to complete the update, and speech denoising can be completed quickly, further improving the efficiency and stability of speech denoising and improving the fluency of speech output.

[0114] In a specific application example, the speech noise reduction method provided in any embodiment of the present application can be deployed in a terminal device applied to scenarios such as video conferencing. In the terminal device, after obtaining the target noise reduction speech processed by the speech noise reduction method, the 3A algorithm is used to process the target noise reduction speech. Among them, the 3A algorithm refers to three audio processing algorithms including acoustic echo cancellation (Acoustic Echo Cancellation, AEC), background noise suppression (Active Noise Control, ANC), and automatic gain control (Automatic Gain Control, AGC). After the speech processed by the 3A algorithm is audio-encoded, it can be transmitted to other terminal devices through the network.

[0115] Figure 6 This is a schematic diagram of the structure of a speech noise reduction system provided in an embodiment of the present application. Figure 6 As shown, the speech noise reduction system 60 includes: a first steering matrix 61, a first blocking matrix 62, a plurality of adaptive matrices 63 and a first noise reduction module 64;

[0116] The first steering matrix 61 is used to suppress the speech signal outside the first target sound pickup area corresponding to the first target sound source in the array speech signal, and obtain a first target speech signal containing a first residual noise signal; the array speech signal is obtained by collecting speech signals in different sound pickup areas by the microphone array;

[0117] The first blocking matrix 62 is used to suppress the voice signal corresponding to the first target pickup area in the array voice signal to obtain a first noise signal;

[0118] The first adaptive matrix 63 is used to convert the first noise signal into a noise signal identical to the first residual noise signal; the first adaptive matrix is ​​obtained by screening a plurality of pre-set adaptive matrices based on the first target sound pickup area;

[0119] The first noise reduction module 64 is used to remove the first residual noise signal in the first target speech signal based on the first noise signal converted by the first adaptive matrix, so as to obtain a first noise-reduced speech signal corresponding to the first target sound source.

[0120] Optionally, the system further includes:

[0121] A second steering matrix corresponding to a second target sound pickup area is used to suppress the speech signal outside the second target sound pickup area in the array speech signal to obtain a second target speech signal containing a second residual noise signal; the second target sound pickup area is each sound pickup area except the first target sound pickup area;

[0122] A second blocking matrix corresponding to the second target sound pickup area is used to suppress the speech signal corresponding to the second target sound pickup area in the array speech signal to obtain a second noise signal;

[0123] A second adaptive matrix is ​​used to convert the second noise signal into a noise signal identical to the second residual noise signal; the second adaptive matrix is ​​obtained by screening a plurality of pre-set adaptive matrices based on the second target sound pickup area;

[0124] The second noise reduction module is used to remove the second residual noise signal in the second target voice signal based on the second noise signal converted by the second adaptive matrix, so as to obtain a second noise-reduced voice signal corresponding to the second target sound pickup area.

[0125] Optionally, it also includes: the number of the adaptive matrices, the total number of the first steering matrices and the second steering matrices, and the total number of the first blocking matrices and the second blocking matrices are the same as the number of pickup areas of the microphone array.

[0126] Optionally, the system further includes:

[0127] An updating module is used to update the weights of the second adaptive matrix based on the second noise reduction speech signal and the second noise signal if the speech signal corresponding to the second target sound pickup area is a non-sound source signal.

[0128] Optionally, the updating of the weights of the second adaptive matrix based on the second noise reduction speech signal and the second noise signal is performed in parallel or in series.

[0129] Optionally, the system further includes:

[0130] A mixing module is used to mix the first noise reduction voice signal and the second noise reduction voice signal to obtain a target noise reduction voice signal of the array voice signal if the voice signal corresponding to the second target sound pickup area is a sound source signal.

[0131] Optionally, the first noise reduction module is specifically used to:

[0132] Using a first multiplier with a gain coefficient of -1 to transform the first noise signal transformed by the first adaptive matrix to obtain a first noise inverse signal;

[0133] An adder is used to add the first anti-noise signal and the first target voice signal to obtain a first noise reduction voice signal corresponding to the first target sound pickup area.

[0134] Optionally, the second noise reduction module is specifically used to:

[0135] Using a second multiplier with a gain coefficient of -1 to transform the second noise signal transformed by the second adaptive matrix to obtain a second noise inverse signal;

[0136] An adder is used to add the second anti-noise signal and the second target voice signal to obtain a second noise reduction voice signal corresponding to the second target sound pickup area.

[0137] The speech noise reduction system provided in the embodiments of the present application can execute the speech noise reduction method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects of the execution method.

[0138] Figure 7A schematic diagram of an electronic device 70 that can be used to implement an embodiment of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present application described and / or required herein.

[0139] like Figure 7 As shown, the electronic device 70 includes at least one processor 71, and a memory connected to the at least one processor 71, such as a read-only memory (ROM) 72, a random access memory (RAM) 73, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 71 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 72 or the computer program loaded from the storage unit 78 to the random access memory (RAM) 73. In the RAM 73, various programs and data required for the operation of the electronic device 70 can also be stored. The processor 71, the ROM 72, and the RAM 73 are connected to each other via a bus 74. An input / output (I / O) interface 75 is also connected to the bus 74.

[0140] A number of components in the electronic device 70 are connected to the I / O interface 75, including: an input unit 76, such as a keyboard, a mouse, etc.; an output unit 77, such as various types of displays, speakers, etc.; a storage unit 78, such as a disk, an optical disk, etc.; and a communication unit 79, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 79 allows the electronic device 70 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0141] The processor 71 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 71 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 71 performs the various methods and processes described above, such as a speech noise reduction method.

[0142] In some embodiments, the speech noise reduction method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 78. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 70 via the ROM 72 and / or the communication unit 79. When the computer program is loaded into the RAM 73 and executed by the processor 71, one or more steps of the speech noise reduction method described above may be performed. Alternatively, in other embodiments, the processor 71 may be configured to perform the speech noise reduction method in any other appropriate manner (e.g., by means of firmware).

[0143] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0144] In some embodiments, the speech noise reduction method may be implemented as a computer program, which is invisibly included in a computer program product. When the computer program is executed by a processor, the speech noise reduction method of the present application is implemented. The computer program product can be understood as a software product that mainly implements its solution through a computer program. The computer program for implementing the method of the present application can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flowchart and / or block diagram are implemented. The computer program can be executed entirely on the machine, partially on the machine, partially on the machine as an independent software package and partially on a remote machine, or entirely on a remote machine or server.

[0145] In the context of the present application, a computer readable storage medium may be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, device or equipment. A computer readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer readable storage medium may be a machine readable signal medium. A more specific example of a machine readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0146] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0147] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0148] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0149] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0150] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A speech noise reduction method, characterized in that: include: Acquire array voice signals of different sound pickup areas collected by the microphone array and obtain a first target sound pickup area corresponding to a first target sound source; Suppressing the speech signal outside the first target sound pickup area in the array speech signal based on the first steering matrix to obtain a first target speech signal containing a first residual noise signal; Removing the speech signal corresponding to the first target sound pickup area from the array speech signal based on the first blocking matrix to obtain a first noise signal; Screening out a first adaptive matrix from a plurality of preset adaptive matrices based on the first target sound pickup area, and converting the first noise signal into a noise signal identical to the first residual noise signal based on the first adaptive matrix; The first residual noise signal in the first target speech signal is removed based on the first noise signal converted by the first adaptive matrix to obtain a first noise reduction speech signal corresponding to the first target sound source.

2. The speech noise reduction method according to claim 1, characterized in that: Also includes: each sound pickup area except the first target sound pickup area is used as a second target sound pickup area; Based on a second steering matrix corresponding to the second target sound pickup area, suppressing the speech signal outside the second target sound pickup area in the array speech signal to obtain a second target speech signal containing a second residual noise signal; Suppressing the speech signal corresponding to the second target sound pickup area in the array speech signal based on the second blocking matrix corresponding to the second target sound pickup area to obtain a second noise signal; Screening out a second adaptive matrix from a plurality of preset adaptive matrices based on the second target sound pickup area, and converting the second noise signal into a noise signal identical to the second residual noise signal based on the second adaptive matrix; The second residual noise signal in the second target speech signal is removed based on the second noise signal converted by the second adaptive matrix to obtain a second noise reduction speech signal corresponding to the second target sound pickup area.

3. The speech noise reduction method according to claim 2, characterized in that: The number of the adaptive matrices, the total number of the first steering matrices and the second steering matrices, and the total number of the first blocking matrices and the second blocking matrices are the same as the number of pickup areas of the microphone array.

4. The speech noise reduction method according to claim 2, characterized in that: Also includes: If the speech signal corresponding to the second target sound pickup area is a non-sound source signal, the weights of the second adaptive matrix are updated based on the second noise reduction speech signal and the second noise signal.

5. The speech noise reduction method according to claim 4, characterized in that: The execution mode of updating the weights of the second adaptive matrix based on the second noise reduction speech signal and the second noise signal is a parallel execution mode or a serial execution mode.

6. The speech noise reduction method according to claim 2, characterized in that: Also includes: If the voice signal corresponding to the second target sound pickup area is a sound source signal, the first noise reduction voice signal and the second noise reduction voice signal are mixed to obtain a target noise reduction voice signal of the array voice signal.

7. The speech noise reduction method according to any one of claims 2 to 6, characterized in that: The step of removing the first residual noise signal in the first target speech signal based on the first noise signal converted by the first adaptive matrix to obtain a first noise-reduced speech signal corresponding to the first target sound source comprises: Using a first multiplier with a gain coefficient of -1 to transform the first noise signal transformed by the first adaptive matrix to obtain a first noise inverse signal; Using an adder to add the first anti-noise signal and the first target voice signal to obtain a first noise-reduced voice signal corresponding to the first target sound pickup area; The step of removing the second residual noise signal in the second target voice signal based on the second noise signal converted by the second adaptive matrix to obtain a second noise-reduced voice signal corresponding to the second target sound pickup area includes: Using a second multiplier with a gain coefficient of -1 to transform the second noise signal transformed by the second adaptive matrix to obtain a second noise inverse signal; An adder is used to add the second anti-noise signal and the second target voice signal to obtain a second noise reduction voice signal corresponding to the second target sound pickup area.

8. A speech noise reduction system, characterized in that: include: A first steering matrix, a first blocking matrix, a plurality of adaptive matrices and a first noise reduction module; wherein, The first steering matrix is ​​used to suppress the speech signal outside the first target sound pickup area corresponding to the first target sound source in the array speech signal to obtain a first target speech signal containing a first residual noise signal; the array speech signal is obtained by collecting speech signals in different sound pickup areas by the microphone array; The first blocking matrix is ​​used to suppress the voice signal corresponding to the first target sound pickup area in the array voice signal to obtain a first noise signal; A first adaptive matrix is ​​used to convert the first noise signal into a noise signal identical to the first residual noise signal; the first adaptive matrix is ​​obtained by screening a plurality of pre-set adaptive matrices based on the first target sound pickup area; The first noise reduction module is used to remove the first residual noise signal in the first target speech signal based on the first noise signal converted by the first adaptive matrix, so as to obtain a first noise-reduced speech signal corresponding to the first target sound source.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the speech noise reduction method according to any one of claims 1 to 7.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program implements the speech noise reduction method according to any one of claims 1 to 7.

Images