Noise reduction method and computer device

Abstract

The present application discloses a noise reduction method and computer device, which belongs to the field of computer technology. The method includes: obtaining n first audio signals collected by n microphones; determining m groups of noise signals in the n first audio signals, each group of noise signals in the m groups of noise signals includes n noise signals; determining m sound source positions according to the m groups of noise signals; for any group of noise signals in the m groups of noise signals, determining signal characteristics according to a group of noise signals and corresponding sound source positions, and generating a denoising signal according to the signal characteristics of a group of noise signals; fusing the n first audio signals and the denoising signal corresponding to each group of noise signals in the m groups of noise signals to obtain a second audio signal. The noise signal characteristics determined by the present application based on the noise signal and its corresponding sound source position can more accurately reflect the characteristics of the noise signal, so the denoising signal of the noise signal characteristics can effectively eliminate the noise signal, thereby improving the noise reduction effect.

Description

Technical Field

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

[0002] With the continuous advancement of science and technology, noise reduction technology has gradually become popular, among which Environmental Noise Cancellation (ENC) technology is a typical example. ENC technology uses multiple microphones to collect audio signals, deeply analyzes the frequency, amplitude, and phase characteristics of the noise, and then generates a signal opposite to the background noise based on the analysis results, thereby achieving noise reduction. In related technologies, time domain analysis is usually used to extract noise signal characteristics. That is, by observing the waveform changes of the noise signal on the time axis and analyzing its amplitude, period, rise time, fall time and other parameters, the characteristics of the noise signal are determined. However, this analysis method often performs poorly in scenarios with more complex noise signals, making it difficult to achieve ideal noise reduction results. Summary of the Invention

[0003] This application provides a noise reduction method and computer device that can improve the noise reduction effect. The technical solution is as follows:

[0004] In a first aspect, a noise reduction method is provided, the method comprising:

[0005] Obtaining n first audio signals, where the n first audio signals are audio signals collected by n microphones, and n is an integer greater than or equal to 2;

[0006] Determine m groups of noise signals from the n first audio signals, each group of noise signals in the m groups of noise signals includes n noise signals, where m is a positive integer;

[0007] Determining m sound source positions according to the m groups of noise signals, wherein the m sound source positions correspond one-to-one to the m groups of noise signals;

[0008] For any one of the m groups of noise signals, determining signal characteristics of the group of noise signals according to the group of noise signals and sound source positions corresponding to the group of noise signals, and generating a denoised signal corresponding to the group of noise signals according to the signal characteristics of the group of noise signals;

[0009] The n first audio signals are fused with a denoised signal corresponding to each noise signal in the m groups of noise signals to obtain a second audio signal.

[0010] In the present application, after the noise reduction system obtains n first audio signals collected by n microphones, it determines m groups of noise signals in the n first audio signals, and each group of noise signals in the m groups of noise signals includes n noise signals. Then, m sound source positions corresponding to the m groups of noise signals are determined. For any group of noise signals in the m groups of noise signals, the signal characteristics of the group of noise signals are determined based on the group of noise signals and their corresponding sound source positions, and the corresponding denoising signal is generated based on the signal characteristics of the group of noise signals. Afterwards, the n first audio signals and the denoising signal corresponding to each group of noise signals in the m groups of noise signals are fused to obtain a second audio signal. In the embodiment of the present application, the noise signal characteristics are determined based on the noise signal and its corresponding sound source position. The noise signal characteristics can more accurately reflect the characteristics of the noise signal in space. In this way, the denoising signal generated based on the noise signal characteristics can effectively eliminate the noise signal, thereby improving the noise reduction effect.

[0011] Optionally, determining m groups of noise signals among the n first audio signals includes:

[0012] For any one of the n first audio signals, input the first audio signal into a first model, and have the first model output a valid signal and m noise signals in the first audio signal;

[0013] All noise signals in the n first audio signals are grouped to obtain the m groups of noise signals.

[0014] Optionally, determining the signal characteristics of the group of noise signals according to the group of noise signals and the sound source positions corresponding to the group of noise signals includes:

[0015] For any one noise signal in the group of noise signals, preprocessing the noise signal and the sound source positions corresponding to the group of noise signals is performed to obtain a group of input data corresponding to the noise signal;

[0016] n groups of input data corresponding to the group of noise signals are input into a second model, and the second model outputs signal features of the group of noise signals.

[0017] Optionally, preprocessing the noise signal and the sound source positions corresponding to the group of noise signals to obtain a group of input data corresponding to the noise signal includes:

[0018] Segmenting the noise signal to obtain multiple audio segments;

[0019] extracting audio features of each audio segment from the plurality of audio segments;

[0020] The audio features of each audio segment in the multiple audio segments are respectively spliced ​​with the position features of the sound source positions corresponding to the set of noise signals to obtain multiple target features, where the multiple target features are a set of input data corresponding to the noise signals.

[0021] Optionally, the second model includes a feature extraction model and a fusion module, and the feature extraction model includes a state space model SSM and a transformer model;

[0022] The SSM is used to output a first signal feature based on the n groups of input data, the transformer model is used to output a second signal feature based on the n groups of input data, and the fusion module is used to output a target signal feature based on the first signal feature and the second signal feature, where the target signal feature is the signal feature of the group of noise signals.

[0023] Optionally, the fusion module is used to: concatenate the first signal feature and the second signal feature to obtain a third signal feature; input the third signal feature into a gating model, and the gating model outputs a first weight, where the first weight is the weight corresponding to the SSM; based on the first weight and the second weight, perform weighted fusion on the first signal feature and the second signal feature to obtain the target signal feature, where the second weight is 1 minus the first weight, and the second weight is the weight corresponding to the transformer model.

[0024] Optionally, after fusing the n first audio signals with the denoised signal corresponding to each noise signal in the m groups of noise signals to obtain the second audio signal, the method further includes:

[0025] fusing all valid signals of the n first audio signals to obtain a third audio signal;

[0026] determining a signal residual between the second audio signal and the third audio signal;

[0027] Parameters in the gating model and the feature extraction model are updated according to the signal residual.

[0028] Optionally, the method further includes:

[0029] Displaying a noise setting interface, wherein the noise setting interface includes multiple sound types;

[0030] In response to receiving a selection operation of at least one sound type from the plurality of sound types, determining the at least one sound type as a noise type;

[0031] According to the noise type set in the noise setting interface, the first model is adjusted so that the first model is used to output a noise signal of the noise type.

[0032] Optionally, the method further includes:

[0033] Displaying a noise reduction mode setting interface, wherein the noise reduction mode setting interface includes one or more noise reduction modes;

[0034] In response to receiving a selection operation of one of the one or more noise reduction modes, determining at least one sound type corresponding to the selected noise reduction mode as a noise type;

[0035] According to the noise type set in the noise reduction mode setting interface, the first model is adjusted so that the first model is used to output a noise signal of the noise type.

[0036] In a second aspect, a noise reduction device is provided, the device comprising:

[0037] An acquisition module, configured to acquire n first audio signals, where the n first audio signals are audio signals collected by n microphones, and n is an integer greater than or equal to 2;

[0038] A first determining module is configured to determine m groups of noise signals among the n first audio signals, where each group of noise signals among the m groups of noise signals includes n noise signals, and m is a positive integer;

[0039] a second determining module, configured to determine m sound source positions according to the m groups of noise signals, wherein the m sound source positions correspond one-to-one to the m groups of noise signals;

[0040] a third determining module, configured to determine, for any one of the m groups of noise signals, signal characteristics of the group of noise signals based on the group of noise signals and sound source locations corresponding to the group of noise signals, and generate a denoised signal corresponding to the group of noise signals based on the signal characteristics of the group of noise signals;

[0041] The first fusion module is configured to fuse the n first audio signals with a denoised signal corresponding to each noise signal in the m groups of noise signals to obtain a second audio signal.

[0042] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the noise reduction method described in the first aspect.

[0043] In a fourth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the noise reduction method described in the first aspect is implemented.

[0044] In a fifth aspect, a computer program product is provided. When the computer program product is run on a computer device, the computer device executes the noise reduction method described in the first aspect.

[0045] It can be understood that the beneficial effects of the second, third, fourth and fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. 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 any creative work.

[0047] Figure 1 is a schematic diagram of a noise reduction system provided in an embodiment of the present application;

[0048] Figure 2 is a diagram of an array structure of a microphone provided by an embodiment of the present application;

[0049] Figure 3 This is a flow chart of a noise reduction method provided in an embodiment of the present application;

[0050] Figure 4 This is a flow chart of a noise recognition process provided by an embodiment of the present application;

[0051] Figure 5 is a flow chart of a noise elimination process provided by an embodiment of the present application;

[0052] Figure 6 Schematic diagram of the structure of a noise reduction device provided in an embodiment of the present application;

[0053] Figure 7 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0055] It should be understood that the “multiple” mentioned in this application refers to two or more. In the description of this application, unless otherwise specified, “ / ” means or, for example, A / B can mean A or B; “and / or” in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in order to facilitate the clear description of the technical solution of this application, words such as “first” and “second” are used to distinguish between identical or similar items with basically the same functions and effects. Those skilled in the art can understand that words such as “first” and “second” do not limit the quantity and execution order, and words such as “first” and “second” do not necessarily limit them to be different.

[0056] The phrases "one embodiment" or "some embodiments" described in this application mean that the specific features, structures, or characteristics described in that embodiment are included in one or more embodiments of the application. Thus, the phrases "in one embodiment," "in some embodiments," "in other embodiments," and "in other embodiments" that appear in different places in this application do not necessarily refer to the same embodiment, but rather mean "one or more but not all embodiments," unless otherwise specifically emphasized. In addition, the terms "including," "comprising," "having," and their variations all mean "including but not limited to," unless otherwise specifically emphasized.

[0057] The application scenarios of the embodiments of the present application are described below.

[0058] The embodiments of the present application are applied in noise reduction scenarios to eliminate ambient noise and improve audio quality. Its application scenarios include but are not limited to teleconferencing, audio recording, and sound monitoring, which require high audio clarity.

[0059] In related technologies, active noise reduction (ANC) or ENC technology is usually used for noise reduction. ANC technology is based on the principle of "interference reduction". It uses a microphone to capture environmental noise, converts the time domain signal into a frequency domain signal, analyzes the frequency distribution of the noise, and extracts the spectral characteristics of the audio signal to obtain the noise characteristics. ENC technology, on the other hand, extracts the noise signal characteristics through time domain analysis. That is, it observes the waveform changes of the noise signal on the time axis, analyzes its amplitude, period, rise time, fall time and other parameters, and then determines the noise characteristics. However, this noise characteristic analysis method is often difficult to accurately analyze the noise characteristics when faced with a constantly changing and complex environment, resulting in the noise reduction effect not being ideal.

[0060] To this end, an embodiment of the present application provides a noise reduction method, which is applied to a noise reduction system. The method collects multiple audio signals through multiple microphones, identifies multiple groups of noise signals therein, and determines the sound source position corresponding to each group of noise signals. Based on the noise signal and its corresponding sound source position, the noise signal characteristics are analyzed, and a corresponding denoising signal is generated. Since the noise signal characteristics are analyzed in combination with the noise signal and its sound source position, it is possible to more accurately distinguish noise sources of different directions and distances, and perform accurate feature analysis on different noise sources, thereby achieving accurate noise reduction. In this way, the method can also achieve good noise reduction effects in complex noise environments. In addition, the embodiment of the present application can also set the noise that the user thinks needs to be eliminated (such as music when sleeping at night, etc.) according to needs. In this way, it can better meet user needs and improve the user's noise reduction experience.

[0061] The noise reduction system provided in the embodiment of the present application is described below.

[0062] Figure 1 Schematic diagram of a noise reduction system provided in an embodiment of the present application. Figure 1 The noise reduction system 10 includes a collection module 101 and a noise reduction module 102 .

[0063] The noise reduction system 10 is used to collect audio signals, which may include various sound signals in the environment. The noise reduction system 10 can reduce the noise of the sound signals and output them. For example, the noise reduction system 10 can be installed in a computer device. For example, the computer device can be a tablet computer, a mobile phone, a recording device, etc., which is not limited in this embodiment of the present application.

[0064] Optionally, the noise reduction system 10 may include a display screen for setting a noise reduction mode of the noise reduction system 10 .

[0065] The acquisition module 101 is used to acquire sound signals. For example, the sound signals acquired by the acquisition module 101 may include noise signals and valid signals other than the noise signals.

[0066] Optionally, the collection module 101 may include multiple sound pickups, each of which is used to collect sound signals from various directions in the environment. For example, the sound pickup may be a microphone.

[0067] Optionally, the multiple pickups in the acquisition module 101 can form a preset array. The preset array can be pre-set. For example, the preset array can be set to a linear array, a circular array, a planar array, a hybrid array, etc., which is not limited in this embodiment of the present application.

[0068] In this case, after the multiple microphones collect sound signals emitted by the same sound source, the noise reduction system 10 determines the location of the sound source according to the time when the multiple microphones collect the sound signals.

[0069] Optionally, the multiple microphones in the acquisition module 101 may include a reference microphone and an error microphone, wherein the reference microphone is used to acquire sound signals in the environment, and the error microphone is used to capture the sound signals after noise reduction.

[0070] For example, Figure 2 As shown, the acquisition module 101 may include four microphones: microphone (MIC) 1, MIC2, MIC3, and MIC4. These four microphones are arranged in a planar array, located at the upper left, lower left, upper right, and lower right corners of the mounting surface, respectively. MIC1, MIC2, and MIC3 are used to acquire sound signals from the environment. After MIC1, MIC2, and MIC3 acquire sound signals from the same sound source, the noise reduction system 10 can determine the location of the sound source based on the time difference between the sound reaching MIC1, MIC2, and MIC3, and then perform noise reduction playback based on this information. MIC4 is used to acquire the played noise-reduced sound signal to evaluate the noise reduction effect.

[0071] After the acquisition module 101 acquires the audio signal, it may send the audio signal to the noise reduction module 102 .

[0072] The noise reduction module 102 is used to reduce noise on the audio signal sent by the acquisition module 101 .

[0073] Optionally, the noise reduction module 102 may include a noise recognition module, a sound preservation module, a feature extraction module, and an anti-phase sound wave generation module.

[0074] The noise recognition module is used to recognize the noise signal in the audio signal and separate the noise signal from the audio signal.

[0075] Optionally, the noise recognition module may include a deep learning algorithm (ie, the first model described below), which is used to recognize a noise signal in the audio signal and then separate the noise signal from the audio signal.

[0076] After the noise recognition module obtains the noise signal in the audio signal, it can send the noise signal to the feature extraction module.

[0077] The sound retaining module is used to retain the audio signal after the noise signal is separated. The audio signal after the noise signal is separated can be called a useful signal (ie, the effective signal described below).

[0078] The feature extraction module may include a feature extraction model and a fusion module. The feature extraction module is used to extract the signal features of the noise signal from the noise signal and send the signal features of the noise signal to the anti-phase sound wave generation module.

[0079] The anti-phase sound wave generating module is used to generate a denoising signal corresponding to the noise signal according to the signal characteristics of the noise signal.

[0080] After the inverse sound wave generating module generates a denoised signal corresponding to the noise signal, the noise reduction module 102 may fuse the audio signal with the denoised signals corresponding to all noise signals in the audio signal to obtain a denoised audio signal.

[0081] Optionally, the noise reduction system 10 may further include a playback module, which is configured to play the noise-reduced audio signal. For example, the playback module may be a speaker.

[0082] The noise reduction method provided in the embodiments of the present application is explained in detail below.

[0083] Figure 3 This is a flow chart of a noise reduction method provided by an embodiment of the present application. This noise reduction method can be applied to the above Figure 2 The noise reduction system 10 in the embodiment, see Figure 3 , the method comprises the following steps:

[0084] Step 301: The noise reduction system obtains n first audio signals, where n is an integer greater than or equal to 2.

[0085] The n first audio signals are audio signals collected by n microphones. For example, the n microphones may be the n microphones in the collection module 101 .

[0086] The first audio signal is an audio signal obtained by the microphones from capturing sounds from the external environment. The n microphones are located in the same environment, and the n first audio signals are captured by the n microphones from the same ambient sound source. The ambient sound source can be a single sound source or a mixed sound source comprising multiple sound sources in the environment. For example, if a user is speaking while a car horn is honking around them, the ambient sound source may include both the user's voice and the car horn.

[0087] It should be noted that within a short period of time, typically between 10 and 30 milliseconds, the characteristics of audio remain essentially unchanged, meaning it is relatively stable, or, in other words, exhibits short-term stability. Therefore, audio processing can be based on a "short-term" basis, meaning the audio can be divided into segments of smaller data volumes for processing, each of which is called a "frame," typically between 10 and 30 milliseconds in length. This not only speeds up processing but also effectively meets real-time requirements. In this embodiment of the present application, the first audio signal can be the audio frame currently being captured by the microphone.

[0088] After acquiring the n first audio signals, the noise reduction system may start to perform noise reduction on the n first audio signals.

[0089] Step 302: The noise reduction system determines m groups of noise signals in the n first audio signals, where each group of noise signals in the m groups of noise signals includes n noise signals, and m is a positive integer.

[0090] The n noise signals in a group of noise signals are noise signals emitted by a sound source collected by the n microphones, that is, the n noise signals are noise signals with relatively similar noise characteristics.

[0091] Since each group of noise signals includes noise signals emitted by the same sound source collected by n microphones, when performing noise reduction subsequently, each group of noise signals in the m groups of noise signals can be targeted for noise reduction. In this way, the noise reduction effect can be significantly improved, ensuring a significant improvement in audio quality.

[0092] In some embodiments, the operation of step 302 may be: for any first audio signal among the n first audio signals, the noise reduction system inputs the first audio signal into the first model, and the first model outputs the valid signal and m noise signals in the first audio signal; and all noise signals in the n first audio signals are grouped to obtain the m groups of noise signals.

[0093] The first model is used to identify and separate noise signals from audio signals. For example, the first model can be a deep learning model. The first model can also be called a noise classifier.

[0094] The effective signal is the audio signal required by the user. For example, the noise reduction system can save the effective signal.

[0095] The noise signal has a sound type. The sound type of the noise signal can be pre-set. For example, the sound type of the noise signal can be set by default or manually set by the user. For example, the sound type of the noise signal can be air conditioning sound, horn sound, engine sound, etc., which is not limited in this embodiment of the present application.

[0096] Since the sound signals emitted by the same sound source are highly similar, the n noise signals with similar time, amplitude, phase and frequency in the n first audio signals can be divided into a group of noise signals. This group of noise signals is emitted by the same sound source.

[0097] In the embodiment of the present application, since the first model is a pre-trained model that can identify the sound types of various sounds, during the noise reduction process of the noise reduction system, when new noise appears in the environment, the first model can accurately identify the noise and perform noise reduction accordingly.

[0098] Optionally, before inputting the first audio signal into the first model, the noise reduction system may preprocess the first audio signal to convert the format of the first audio signal into a format suitable for input of the first model, so as to improve the accuracy of identifying the noise signal.

[0099] For ease of understanding, the following Figure 4 The above noise recognition process is exemplified.

[0100] Figure 4 This is a flow chart of a noise identification method provided in an embodiment of the present application, see Figure 4 The method may include steps 401 to 405:

[0101] Step 401: The noise reduction system obtains n first audio signals, where the n first audio signals are audio signals collected by n microphones.

[0102] Step 402: The noise reduction system pre-processes the n first audio signals to obtain n processed first audio signals.

[0103] Step 403: The noise reduction system inputs the n first audio signals into a first model respectively.

[0104] Step 404: For any one of the n first audio signals, the first model performs feature extraction on the first audio signal to obtain one or more audio features.

[0105] Step 405: The first model identifies one or more noise features among the one or more audio features, separates one or more noise signals corresponding to the one or more noise features from the first audio signal, and outputs a valid signal remaining after the noise signals are separated and the one or more noise signals.

[0106] In some embodiments, after the first model outputs the valid signal, the noise reduction system may subsequently display the valid signal on a noise settings interface. This noise settings interface includes one or more sound types corresponding to one or more sound signals in the valid signal. Users can set the noise type to be eliminated in this noise settings interface. This customized noise type approach can better meet the user's personalized needs, improve the user experience, and thus enhance the usability of the noise reduction system.

[0107] The following describes the process of setting the noise type.

[0108] The noise type setting process may include the following steps (1) to (3):

[0109] Step (1): The noise reduction system displays a noise setting interface, which includes multiple sound types.

[0110] For example, the multiple sound types may include a sound type corresponding to each sound signal in the multiple sound signals in the valid signal, and / or the multiple sound types may include some sound types preset by a technician.

[0111] After the noise reduction system displays the noise setting interface, the user can select the sound type to be set in the noise setting interface to set the noise type.

[0112] Step (2): In response to receiving a selection operation for at least one sound type among the multiple sound types, the noise reduction system determines the at least one sound type as a noise type.

[0113] For example, the user may click on at least one sound type in the noise setting interface. In response to the click operation, the noise reduction system may display the at least one sound type as a selected state.

[0114] Step (3): The noise reduction system adjusts the first model according to the noise type set in the noise setting interface, so that the first model is used to output a noise signal of the noise type.

[0115] It should be noted that during the model training phase, the first model can identify and separate noise signals from multiple noise types. During the model deployment phase, the noise reduction system can flexibly adjust the noise types that the first model can separate based on actual needs (such as user settings in the noise settings interface), thereby better meeting noise reduction needs.

[0116] Optionally, the noise reduction system further includes one or more noise reduction modes, each of the one or more noise reduction modes corresponds to at least one sound type, and all sound types corresponding to a noise reduction mode are noise types that need to be eliminated in this noise reduction mode.

[0117] The one or more noise reduction modes can be pre-set. For example, the one or more noise reduction modes can be set according to demand. For example, the one or more noise reduction modes can include office mode, home mode, outdoor mode, etc., which is not limited in the embodiments of the present application.

[0118] In this way, users can choose different noise reduction modes according to different scenarios, which can better meet user needs and improve the user's noise reduction experience.

[0119] The following describes the process of setting the noise reduction mode.

[0120] The noise reduction mode setting process may include the following steps (1) to (3):

[0121] Step (1): The noise reduction system displays a noise reduction mode setting interface, which includes one or more noise reduction modes.

[0122] After the noise reduction system displays the noise reduction mode setting interface, the user can select a noise reduction mode to set on the noise reduction mode setting interface.

[0123] Step (2): In response to receiving a selection operation for one of the one or more noise reduction modes, the noise reduction system determines at least one sound type corresponding to the selected noise reduction mode as a noise type.

[0124] For example, the user may click on a noise reduction mode in the noise reduction mode setting interface. In response to the click operation, the noise reduction system may display the noise reduction mode as a selected state.

[0125] Step (3): The noise reduction system adjusts the first model according to the noise type set in the noise reduction mode setting interface, so that the first model is used to output a noise signal of the noise type.

[0126] It should be noted that during the model training phase, the first model can identify and separate noise signals from multiple noise types. During the model's operational phase, the noise reduction system can flexibly adjust the noise types that the first model can separate based on actual needs (such as user settings in the noise reduction mode settings interface), thereby better meeting noise reduction needs.

[0127] Step 303: The noise reduction system determines m sound source positions according to the m groups of noise signals, where the m sound source positions correspond one-to-one to the m groups of noise signals.

[0128] For example, for any one of the m groups of noise signals, the noise reduction system can determine the time delay difference of arrival (TDOA) of the signal acquisition based on the time instants when the n microphones acquire the n noise signals in the group. Based on the TDOA, the system can then determine the direction of arrival (DOA) of the sound source corresponding to the group of noise signals relative to the n microphones. The TDOA and the DOA represent the location of the sound source corresponding to the group of noise signals. For example, the noise reduction system can determine the TDOA using a generalized cross-correlation phase transform (GCC-PHAT) algorithm.

[0129] For example, the noise reduction system can determine the direction of arrival according to the delay difference using the following formula:

[0130] ;

[0131] in, is the wave direction; is the speed of sound (about 340 m / s); is the delay difference; is the spacing between pickups.

[0132] After determining the sound source position corresponding to each of the m groups of noise signals, the noise reduction system can subsequently generate a more accurate denoising signal based on the sound source position corresponding to the noise signal. Moreover, when the sound source position corresponding to the noise signal changes, the noise reduction system can determine the changed sound source position in real time, thereby ensuring the stability and accuracy of noise reduction.

[0133] Step 304: For any one of the m groups of noise signals, the noise reduction system determines the signal characteristics of the group of noise signals based on the group of noise signals and the sound source positions corresponding to the group of noise signals, and generates a denoised signal corresponding to the group of noise signals based on the signal characteristics of the group of noise signals.

[0134] For example, the noise reduction system can generate a denoised signal corresponding to the set of noise signals using an inverse sound wave generation module. For example, the noise reduction system can restore the signal characteristics of the set of noise signals into a spectrogram, then convert the spectrogram into a time-domain signal that is the opposite of the noise signal (e.g., through an inverse short-time Fourier transform), and finally perform amplitude normalization on the time-domain signal to obtain the denoised signal corresponding to the set of noise signals.

[0135] Since sound is affected by distance, direction and environmental factors during its propagation, and the location of the sound source can reflect important information such as the sound propagation path and environment, the signal characteristics of the noise signal can be determined more accurately based on the noise signal and the sound source location corresponding to the noise signal. Subsequently, after generating a denoising signal based on the signal characteristics, the denoising signal can effectively eliminate the noise signal without affecting other sound signals, thereby improving the noise reduction effect.

[0136] It should be noted that in the embodiments of the present application, when generating the denoising signal, the inverted sound wave generation module can automatically adjust parameters based on the current environment, such as adjusting the frequency filtering range and enhancing the noise suppression strength of specific frequency bands, to ensure that good noise reduction effects can be achieved in scenarios with different noise variations. For example, in a noisy urban environment, the focus can be on suppressing low-frequency engine sounds, while in a quieter indoor environment, the processing of high-frequency noise can be reduced to maintain the natural and clear audio after noise reduction.

[0137] In some embodiments, the operation of the noise reduction system determining the signal characteristics of the set of noise signals based on the set of noise signals and the sound source positions corresponding to the set of noise signals may include the following steps (1) to (2):

[0138] Step (1): For any one noise signal in the group of noise signals, the noise reduction system preprocesses the noise signal and the sound source position corresponding to the group of noise signals to obtain a group of input data corresponding to the noise signal.

[0139] Preprocessing the noise signal and the sound source positions corresponding to the group of noise signals refers to processing the noise signal and the sound source positions corresponding to the group of noise signals into input data suitable for the second model.

[0140] By preprocessing a set of noise signals and their corresponding sound source positions, the format of the set of noise signals and their corresponding sound source positions can be unified and standardized so as to meet the input standard of the second model, thereby ensuring the accuracy of the signal characteristics output by the second model.

[0141] For example, the noise reduction system can segment the noise signal to obtain multiple audio segments, then extract the audio features of each audio segment in the multiple audio segments, and then splice the audio features of each audio segment in the multiple audio segments with the position features of the sound source position corresponding to the group of noise signals to obtain multiple target features, which are a group of input data corresponding to the noise signal.

[0142] It should be noted that since the noise signal is continuous in time, the multiple audio segments obtained by segmenting it are sequential in time, forming a time series. The multiple target features obtained by concatenating each of the multiple audio segments with the positional features corresponding to the sound source position of the noise signal also maintain temporal sequence, forming a time series.

[0143] After the noise reduction system preprocesses the set of noise signals and the sound source positions corresponding to the set of noise signals to obtain n sets of input data corresponding to the set of noise signals, the n sets of input data can be input into the second model to obtain the signal characteristics of the set of noise signals.

[0144] Step (2): The noise reduction system inputs n groups of input data corresponding to the group of noise signals into the second model, and the second model outputs signal features of the group of noise signals.

[0145] For example, the noise reduction system inputs the n groups of input data into the n channels in the second model one by one. The second model can perform feature extraction on the input data of each of the n channels to obtain m signal features, and then fuse the n signal features across channels to obtain the signal features of the group of noise signals.

[0146] Optionally, the second model includes a feature extraction model and a fusion module, wherein the feature extraction model includes a state space model (SSM) and a transformer model. Optionally, both the SSM and the transformer model may have n channels.

[0147] The SSM is configured to output a first signal feature based on the n sets of input data, the transformer model is configured to output a second signal feature based on the n sets of input data, and the fusion module is configured to output a target signal feature based on the first and second signal features. The target signal feature is the signal feature of the set of noise signals.

[0148] For example, after the noise reduction system inputs the n sets of input data into the second model, the SSM in the second model can perform feature extraction on each of the n sets of input data to obtain n signal features, then perform feature fusion on the n signal features to obtain a first signal feature, and send the first signal feature to the fusion module; the transformer model in the second model can perform feature extraction on each of the n sets of input data to obtain n signal features, then perform feature fusion on the n signal features to obtain a second signal feature, and send the second signal feature to the fusion module. The fusion module can perform feature fusion on the first signal feature and the second signal feature to obtain a target signal feature.

[0149] For example, the SSM can determine the signal characteristics by the following formula:

[0150] ;

[0151] ;

[0152] in, represents the state vector of the SSM at time t; is the state transfer matrix; represents the state vector of the SSM at time t-1; is the input matrix; For input data; is the signal feature; is the observation matrix; is the observation input matrix.

[0153] For example, the transformer model can determine the signal characteristics through the following formula:

[0154] ;

[0155] in, is the signal feature; is the query vector; is the key vector; is a value vector; is the dimension scaling factor; To perform the transpose operation on the K vector.

[0156] The SSM in the second model can efficiently capture the short-term dependencies of stationary signals, which enables it to quickly adapt to signal changes when processing sudden changes in signals. When noise mutations occur, the SSM can accurately generate the signal features of the input data. The computational complexity of the SSM is linear, which makes it more efficient when processing large-scale input data. The transformer model in the second model can capture long-distance dependencies through the self-attention mechanism, thereby performing well in processing audio signals with global contextual information. This enables the transformer model to extract signal features more accurately when the noise is relatively stable. In this way, whether the noise is stable or the noise mutation occurs, the second model can output more accurate signal features, thereby significantly improving the noise reduction effect.

[0157] In some embodiments, the fusion module can concatenate the first signal feature and the second signal feature to obtain a third signal feature; input the third signal feature into a gating model, and the gating model outputs a first weight; based on the first weight and the second weight, perform weighted fusion on the first signal feature and the second signal feature to obtain a target signal feature, where the second weight is 1 minus the first weight.

[0158] This gating model can dynamically adjust the weights according to the changes in the noise signal, so that the fused target signal features can better adapt to the scenarios of noise mutation and stable noise.

[0159] The first weight is the weight corresponding to the SSM, and the second weight is the weight corresponding to the transformer model.

[0160] The gating model can adaptively adjust the weights of the signal features output by the SSM and transformer models. In this way, in the case of a sudden change in noise, the weight of the first signal feature can be made higher so that the target signal feature can better highlight the signal feature when the noise suddenly changes; in the case of stable noise, the weight of the second signal feature can be made higher so that the target signal feature can better highlight the signal feature when the noise is stable. By dynamically adjusting the weights of the signal features, the fusion module can better handle the situations of sudden noise changes and stable noise, thus improving the accuracy of the output target signal features and improving the noise reduction effect.

[0161] Step 305: The noise reduction system fuses the n first audio signals with the denoised signal corresponding to each noise signal in the m groups of noise signals to obtain a second audio signal.

[0162] The second audio signal is an audio signal obtained by performing noise reduction on the n first audio signals.

[0163] Since the m denoising signals can eliminate the m groups of noise signals, the n first audio signals and the denoising signal corresponding to each noise signal in the m groups of noise signals can be fused to obtain a noise-reduced audio signal (i.e., a second audio signal).

[0164] For example, the noise reduction system may combine the n first audio signals to obtain a fourth audio signal, and then, based on a time alignment principle, superimpose the fourth audio signal and the m denoised signals to obtain a second audio signal.

[0165] Optionally, after obtaining the second audio signal, the noise reduction system may play the second audio signal; alternatively, the noise reduction system may save the second audio signal for recording; alternatively, the noise reduction system may also send the second audio signal to other devices.

[0166] In some embodiments, after obtaining the second audio signal, the noise reduction system may fuse all valid signals in the n first audio signals to obtain a third audio signal; determine a signal residual between the second audio signal and the third audio signal; and update parameters in the gating model and the feature extraction model based on the signal residual.

[0167] The third audio signal is an audio signal after noise reduction in an ideal state.

[0168] The signal residual is used to evaluate the noise reduction effect. Specifically, a smaller signal residual indicates a smaller difference between the third audio signal and the second audio signal, i.e., a better noise reduction effect. A larger signal residual indicates a larger difference between the third audio signal and the second audio signal, i.e., a worse noise reduction effect. Therefore, the parameters in the gating model and the feature extraction model can be updated based on the signal residual to optimize the noise reduction capability of the noise reduction system.

[0169] In this way, if a new noise source appears in the environment, or if existing noise suddenly changes, the noise reduction system's noise reduction parameters can be continuously adjusted, allowing the system to adapt to the current environment. This allows it to more accurately identify the signal characteristics of the noise signal in the environment, thereby ensuring the noise reduction capability of the noise reduction system. For example, if there is a sudden increase in traffic noise or the start-up of some mechanical equipment, the noise reduction system can quickly identify this change and adjust the noise reduction parameters accordingly. This allows the noise reduction system's noise reduction capabilities to quickly adapt to the current environment, accurately determining the signal characteristics of the noise signal in the current environment.

[0170] For example, the noise reduction system may determine the signal residual according to the second audio signal and the third audio signal using the following formula:

[0171] ;

[0172] in, is the signal residual; is a second audio signal; is the third audio signal.

[0173] For example, the noise reduction system can determine a loss value based on the signal residual through a loss function (such as a mean square error function), and then use a backpropagation algorithm to update the parameters in the gating model and the feature extraction model respectively based on the loss value.

[0174] For ease of understanding, the following Figure 5 The above noise elimination process is exemplified.

[0175] Figure 5 This is a schematic diagram of a noise elimination process provided in an embodiment of the present application.

[0176] like Figure 5 As shown, the noise reduction system inputs multiple sets of input data into a feature extraction model, which extracts features from the multiple sets of input data and outputs a first signal feature and a second signal feature. The fusion module fuses the first signal feature and the second signal feature to obtain a target signal feature. The inverse sound wave generation module generates a denoised signal based on the target signal feature. The denoised signal fusion module fuses the first audio signal with the denoised signal to obtain a second audio signal. The feedback controller determines a signal residual based on the second audio signal, updates the parameters in the fusion module based on the signal residual, and updates the parameters in the feature extraction model based on the signal residual.

[0177] In an embodiment of the present application, after the noise reduction system obtains n first audio signals collected by n microphones, it determines m groups of noise signals in the n first audio signals, and each group of noise signals in the m groups of noise signals includes n noise signals. Then, m sound source positions corresponding to the m groups of noise signals are determined. For any group of noise signals in the m groups of noise signals, the signal characteristics of the group of noise signals are determined based on the group of noise signals and their corresponding sound source positions, and a corresponding denoising signal is generated based on the signal characteristics of the group of noise signals. Thereafter, the n first audio signals and the denoising signal corresponding to each group of noise signals in the m groups of noise signals are fused to obtain a second audio signal. In the embodiment of the present application, the noise signal characteristics are determined based on the noise signal and its corresponding sound source position. The noise signal characteristics can more accurately reflect the characteristics of the noise signal in space. In this way, the denoising signal generated based on the noise signal characteristics can effectively eliminate the noise signal, thereby improving the noise reduction effect.

[0178] Figure 6This is a schematic diagram of the structure of a noise reduction device provided by an embodiment of the present application. The device can be implemented as part or all of a computer device by software, hardware, or a combination of both. The computer device can be the following Figure 7 Computer equipment shown. Figure 6 The device includes: an acquisition module 601, a first determination module 602, a second determination module 603, a third determination module 604, and a first fusion module 605.

[0179] An acquisition module 601 is configured to acquire n first audio signals, where the n first audio signals are audio signals collected by n microphones, and n is an integer greater than or equal to 2;

[0180] A first determining module 602 is configured to determine m groups of noise signals from the n first audio signals, where each group of noise signals in the m groups includes n noise signals, and m is a positive integer;

[0181] A second determining module 603 is configured to determine m sound source positions according to the m groups of noise signals, where the m sound source positions correspond one-to-one to the m groups of noise signals;

[0182] a third determining module 604 configured to determine, for any one of the m groups of noise signals, signal characteristics of the group of noise signals based on the group of noise signals and sound source locations corresponding to the group of noise signals, and generate a denoised signal corresponding to the group of noise signals based on the signal characteristics of the group of noise signals;

[0183] The first fusion module 605 is configured to fuse the n first audio signals with the denoised signal corresponding to each noise signal in the m groups of noise signals to obtain a second audio signal.

[0184] Optionally, the first determining module 602 is configured to:

[0185] For any one of the n first audio signals, input the first audio signal into a first model, and the first model outputs a valid signal in the first audio signal and m noise signals;

[0186] All noise signals in the n first audio signals are grouped to obtain the m groups of noise signals.

[0187] Optionally, the third determining module 604 is configured to:

[0188] For any one noise signal in the set of noise signals, preprocessing the noise signal and the sound source positions corresponding to the set of noise signals is performed to obtain a set of input data corresponding to the noise signal;

[0189] N groups of input data corresponding to the group of noise signals are input into the second model, and the second model outputs signal features of the group of noise signals.

[0190] Optionally, the third determining module 604 is configured to:

[0191] Segmenting the noise signal to obtain multiple audio segments;

[0192] extracting audio features of each audio segment from the plurality of audio segments;

[0193] The audio features of each audio segment in the multiple audio segments are respectively spliced ​​with the position features of the sound source positions corresponding to the group of noise signals to obtain multiple target features, where the multiple target features are a group of input data corresponding to the noise signal.

[0194] Optionally, the second model includes a feature extraction model and a fusion module, the feature extraction model includes a state space model SSM and a transformer model;

[0195] The SSM is used to output a first signal feature based on the n sets of input data, the transformer model is used to output a second signal feature based on the n sets of input data, and the fusion module is used to output a target signal feature based on the first signal feature and the second signal feature, where the target signal feature is the signal feature of the set of noise signals.

[0196] Optionally, the fusion module is used to: concatenate the first signal feature and the second signal feature to obtain a third signal feature; input the third signal feature into a gating model, and the gating model outputs a first weight, where the first weight is the weight corresponding to the SSM; based on the first weight and the second weight, perform weighted fusion on the first signal feature and the second signal feature to obtain the target signal feature, where the second weight is 1 minus the first weight, and the second weight is the weight corresponding to the transformer model.

[0197] Optionally, the device further comprises:

[0198] a second fusion module, configured to fuse all valid signals in the n first audio signals to obtain a third audio signal;

[0199] a fourth determining module, configured to determine a signal residual between the second audio signal and the third audio signal;

[0200] An updating module is used to update parameters in the gating model and the feature extraction model according to the signal residual.

[0201] Optionally, the device further comprises:

[0202] A first display module is used to display a noise setting interface, which includes multiple sound types;

[0203] a fifth determining module, configured to, in response to receiving a selection operation on at least one sound type among the multiple sound types, determine the at least one sound type as a noise type;

[0204] The first adjustment module is used to adjust the first model according to the noise type set in the noise setting interface, so that the first model is used to output a noise signal of the noise type.

[0205] Optionally, the device further comprises:

[0206] A second display module is used to display a noise reduction mode setting interface, where the noise reduction mode setting interface includes one or more noise reduction modes;

[0207] a sixth determining module, configured to, in response to receiving a selection operation of one of the one or more noise reduction modes, determine at least one sound type corresponding to the selected noise reduction mode as a noise type;

[0208] The second adjustment module is used to adjust the first model according to the noise type set in the noise reduction mode setting interface, so that the first model is used to output a noise signal of the noise type.

[0209] In an embodiment of the present application, after the noise reduction system obtains n first audio signals collected by n microphones, it determines m groups of noise signals in the n first audio signals, and each group of noise signals in the m groups of noise signals includes n noise signals. Then, m sound source positions corresponding to the m groups of noise signals are determined. For any group of noise signals in the m groups of noise signals, the signal characteristics of the group of noise signals are determined based on the group of noise signals and their corresponding sound source positions, and a corresponding denoising signal is generated based on the signal characteristics of the group of noise signals. Thereafter, the n first audio signals and the denoising signal corresponding to each group of noise signals in the m groups of noise signals are fused to obtain a second audio signal. In the embodiment of the present application, the noise signal characteristics are determined based on the noise signal and its corresponding sound source position. The noise signal characteristics can more accurately reflect the characteristics of the noise signal in space. In this way, the denoising signal generated based on the noise signal characteristics can effectively eliminate the noise signal, thereby improving the noise reduction effect.

[0210] It should be noted that: the noise reduction device provided in the above embodiment only uses the division of the above-mentioned functional modules as an example to illustrate the noise reduction. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0211] The functional modules in the above embodiments may be integrated into a single processing unit, or each functional module may exist physically as a separate processing unit, or two or more functional modules may be integrated into a single processing unit. The above processing unit may be implemented in either hardware or software. Furthermore, the specific names of the functional modules are merely for the purpose of distinguishing them from one another and are not intended to limit the scope of protection of the embodiments of this application.

[0212] The noise reduction device and noise reduction method provided in the above embodiments belong to the same concept. The specific working process of the functional modules in the above embodiments and the technical effects brought about can be found in the method embodiment part and will not be repeated here.

[0213] Figure 7 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. Figure 7 As shown, the computer device 7 includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and executable on the processor 70. When the processor 70 executes the computer program 72, the steps of the noise reduction method in the above embodiment are implemented.

[0214] The computer device 7 can be a general-purpose computer device or a dedicated computer device. In a specific implementation, the computer device 7 can be a desktop computer, a portable computer, a network server, a palmtop computer, a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiment of the present application does not limit the type of the computer device 7. Those skilled in the art will understand that Figure 7 This is merely an example of the computer device 7 and does not constitute a limitation on the computer device 7 . The computer device 7 may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device 7 may also include input and output devices, network access devices, etc.

[0215] The processor 70 may be a central processing unit (CPU), or may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor may be a microprocessor or any conventional processor.

[0216] In some embodiments, the memory 71 may be an internal storage unit of the computer device 7, such as the hard drive or memory of the computer device 7. In other embodiments, the memory 71 may also be an external storage device of the computer device 7, such as a plug-in hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the computer device 7. Furthermore, the memory 71 may include both the internal storage unit of the computer device 7 and an external storage device. The memory 71 is used to store the operating system, application programs, boot loader, data, and other programs. The memory 71 may also be used to temporarily store data that has been output or is about to be output.

[0217] An embodiment of the present application also provides a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor implements the steps of any of the above-mentioned method embodiments when executing the computer program.

[0218] An embodiment of the present application further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0219] An embodiment of the present application provides a computer program product, which, when executed on a computer, enables the computer to execute the steps in the above-mentioned various method embodiments.

[0220] If the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the processes in the above-mentioned method embodiments by using a computer program to instruct the relevant hardware. This computer program can be stored in a computer-readable storage medium. When executed by a processor, this computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, an executable file, or some intermediate form. The computer-readable medium can include at least any entity or device capable of carrying the computer program code to a camera / terminal device, a recording medium, computer memory, ROM (read-only memory), RAM (random access memory), CD-ROM (compact disc read-only memory), magnetic tape, floppy disk, and optical data storage device. The computer-readable storage medium mentioned in this application can be a non-volatile storage medium, in other words, a non-transitory storage medium.

[0221] It should be understood that all or part of the steps for implementing the above embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the steps may be implemented in the form of a computer program product. The computer program product may include one or more computer instructions. The computer instructions may be stored in the above-mentioned computer-readable storage medium.

[0222] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0223] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0224] In the embodiments provided in this application, it should be understood that the disclosed apparatus / computer equipment and methods can be implemented in other ways. For example, the apparatus / computer equipment embodiments described above are merely schematic. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of the apparatus or unit, which can be electrical, mechanical or other forms.

[0225] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0226] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant regulations and standards of relevant countries and regions, and corresponding operation entrances are provided for users to choose to authorize or refuse.

[0227] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A noise reduction method, characterized in that: The method comprises: Obtaining n first audio signals, where the n first audio signals are audio signals collected by n microphones, and n is an integer greater than or equal to 2; Determine m groups of noise signals from the n first audio signals, each group of noise signals in the m groups of noise signals includes n noise signals, where m is a positive integer; Determining m sound source positions according to the m groups of noise signals, wherein the m sound source positions correspond one-to-one to the m groups of noise signals; For any one noise signal in any one group of noise signals in the m groups of noise signals, the noise signal is segmented to obtain multiple audio segments; audio features of each audio segment in the multiple audio segments are extracted; the audio features of each audio segment in the multiple audio segments are respectively spliced ​​with the position features of the sound source position corresponding to the group of noise signals to obtain multiple target features, and the multiple target features are a group of input data corresponding to the noise signal; n groups of input data corresponding to the group of noise signals are input into a second model, the second model outputs the signal features of the group of noise signals, and a denoised signal corresponding to the group of noise signals is generated according to the signal features of the group of noise signals; wherein the second model includes a feature extraction model and a fusion module, the feature extraction model includes a state space model SSM and a transformer model; the SSM is used to output a first signal feature according to the n groups of input data, the transformer model is used to output a second signal feature according to the n groups of input data, and the fusion module is used to output a target signal feature according to the first signal feature and the second signal feature, and the target signal feature is the signal feature of the group of noise signals; The n first audio signals are fused with a denoised signal corresponding to each noise signal in the m groups of noise signals to obtain a second audio signal.

2. The method according to claim 1, wherein The determining of m groups of noise signals among the n first audio signals includes: For any one of the n first audio signals, input the first audio signal into a first model, and have the first model output a valid signal and m noise signals in the first audio signal; All noise signals in the n first audio signals are grouped to obtain the m groups of noise signals.

3. The method according to claim 1 or 2, wherein: The fusion module is used to: splice the first signal feature and the second signal feature to obtain a third signal feature; input the third signal feature into a gating model, and the gating model outputs a first weight, where the first weight is the weight corresponding to the SSM; Based on the first weight and the second weight, the first signal feature and the second signal feature are weightedly fused to obtain the target signal feature, where the second weight is obtained by subtracting the first weight from 1, and the second weight is the weight corresponding to the transformer model.

4. The method according to claim 3, wherein After fusing the n first audio signals with the denoised signal corresponding to each noise signal in the m groups of noise signals to obtain a second audio signal, the method further includes: fusing all valid signals of the n first audio signals to obtain a third audio signal; determining a signal residual between the second audio signal and the third audio signal; Parameters in the gating model and the feature extraction model are updated according to the signal residual.

5. The method according to claim 2, wherein The method further comprises: Displaying a noise setting interface, wherein the noise setting interface includes multiple sound types; In response to receiving a selection operation of at least one sound type from the plurality of sound types, determining the at least one sound type as a noise type; According to the noise type set in the noise setting interface, the first model is adjusted so that the first model is used to output a noise signal of the noise type.

6. The method according to claim 2, wherein The method further comprises: Displaying a noise reduction mode setting interface, wherein the noise reduction mode setting interface includes one or more noise reduction modes; In response to receiving a selection operation of one of the one or more noise reduction modes, determining at least one sound type corresponding to the selected noise reduction mode as a noise type; According to the noise type set in the noise reduction mode setting interface, the first model is adjusted so that the first model is used to output a noise signal of the noise type.

7. A computer device, characterized in that: The computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the method according to any one of claims 1 to 6 when executed by the processor.

Images