Audio signal processing method, computer equipment and storage medium

Abstract

The invention provides an audio signal processing method, computer equipment and a storage medium, and the method comprises the steps: obtaining an initial audio signal; determining a target filter coefficient corresponding to at least one target sound area based on the filter coefficients corresponding to the plurality of sound areas, the filter coefficients being obtained by training impulse response data received by audio input devices in the plurality of sound areas by a neural network model, the target sound area is one or more sound areas in the plurality of sound areas; according to the target filter coefficient corresponding to the at least one target sound area, filtering the initial audio signal to obtain at least one target audio signal; and transmitting each target audio signal to the corresponding target sound area. Therefore, the effect of audio partition playing can be achieved, and the intelligence of audio output is improved.

Description

Technical Field

[0001] The present application relates to the technical field of sound field control, and in particular to an audio signal processing method, a computer device and a storage medium. Background Art

[0002] With the development of the times, people's personalized requirements for audio systems are gradually increasing in many scenarios. For example, take people's choice of audio while riding in a car. The front passengers may want to listen to navigation voice, and the back passengers may want to listen to pop music. The choice of audio is different. However, the traditional car audio system cannot meet the personalized needs of everyone at the same time, resulting in some passengers having a poor experience and the audio output is not smart enough.

[0003] Therefore, how to improve the intelligence of audio output becomes a problem that needs to be solved urgently. Summary of the invention

[0004] The present application provides an audio signal processing method, a computer device and a storage medium, which can improve the intelligence of audio output.

[0005] In a first aspect, the present application provides an audio signal processing method, the method comprising:

[0006] Obtaining an initial audio signal;

[0007] Determine a target filter coefficient corresponding to at least one target sound zone based on filter coefficients corresponding to the plurality of sound zones, wherein the filter coefficient is obtained by training a neural network model based on impulse response data received by an audio input device in the plurality of sound zones, and the target sound zone is one or more sound zones in the plurality of sound zones;

[0008] According to the target filter coefficient corresponding to at least one target sound zone, the initial audio signal is filtered to obtain at least one target audio signal;

[0009] Each of the target audio signals is transmitted to the corresponding target sound zone.

[0010] In a second aspect, the present application further provides a computer device, the computer device comprising:

[0011] Memory and processor;

[0012] Wherein, the memory is connected to the processor and is used to store programs;

[0013] The processor is used to implement the steps of any one of the audio signal processing methods provided in the embodiments of the present application by running the program stored in the memory.

[0014] In a third aspect, 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 processor implements the steps of the audio signal processing method as described in any one of the embodiments of the present application.

[0015] The audio signal processing method, computer device and storage medium disclosed in the present application, after obtaining the initial audio signal, determine the target filter coefficient corresponding to at least one target sound zone (the target sound zone is one or more sound zones among the multiple sound zones) based on the filter coefficients corresponding to multiple sound zones (the filter coefficients are obtained by training the neural network model based on the impulse response data received by the audio input devices in the multiple sound zones), and filter the initial audio signal according to the target filter coefficient corresponding to the at least one target sound zone to obtain at least one target audio signal, and then transmit each target audio signal to the corresponding target sound zone, thereby achieving the effect of audio partition playback and improving the intelligence of audio output.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of 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 some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 is a schematic diagram of the steps of an audio signal processing method provided in an embodiment of the present application;

[0019] Figure 2 It is a schematic diagram of the steps of training a neural network model provided in an embodiment of the present application;

[0020] Figure 3 It is a schematic diagram of obtaining filter coefficients based on a neural network model provided in an embodiment of the present application;

[0021] Figure 4 is a schematic diagram of a step of performing model training on the neural network model according to the plurality of impulse response data provided by an embodiment of the present application;

[0022] Figure 5 It is a schematic diagram of inputting amplitude spectrum data and phase spectrum data into different branch modules for processing, provided by an embodiment of the present application;

[0023] Figure 6is a schematic diagram of an audio output process provided by an embodiment of the present application;

[0024] Figure 7 It is a schematic block diagram of a computer device provided in an embodiment of the present application.

[0025] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are 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 are within the scope of protection of this application.

[0027] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0028] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0029] It should be understood that in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first", "second" and the like are used to distinguish between the same items or similar items with substantially the same functions and effects. For example, the first callback function and the second callback function are only used to distinguish different callback functions, and their order is not limited. Those skilled in the art can understand that the words "first", "second" and the like do not limit the quantity and execution order, and the words "first", "second" and the like do not necessarily limit them to be different.

[0030] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0031] Nowadays, people's personalized requirements for audio systems are gradually increasing in many scenarios. For example, when people are riding in a car, the front passengers may want to listen to navigation voice, while the back passengers may want to listen to pop music. The audio choices are different. However, the traditional in-car audio system cannot meet the personalized needs of everyone at the same time, resulting in some passengers having a poor experience and the audio output is not smart enough.

[0032] To this end, embodiments of the present application provide an audio signal processing method, a computer device, and a storage medium to improve the intelligence of audio output.

[0033] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0034] See also Figure 1 , Figure 1 The present invention is a flowchart of an audio signal processing method provided in an embodiment of the present invention. The audio signal processing method can be applied in a terminal device, wherein the terminal device can be a mobile phone, a tablet computer, a laptop computer, a desktop computer, a personal digital assistant, a wearable device or a smart screen, a smart TV, a projector, a reading pen, a computer device, etc., but is not limited thereto.

[0035] like Figure 1 As shown, the audio signal processing method can be applied to a computer device, and the audio signal processing method applied to a computer device includes steps S101 to S104.

[0036] S101: Acquire an initial audio signal.

[0037] The initial audio signal refers to the audio signal that the audio output device in the current scene originally wants to output. It should be noted that in actual applications, some of the multiple users in the current scene may want to hear the audio, while some do not. In order to allow users who want to hear the audio to receive the audio and users who do not want to hear the audio to not be disturbed by the audio, each user can enjoy a personalized listening experience according to their own needs. The audio output device does not directly output the initial audio signal. The audio output device includes but is not limited to a speaker.

[0038] For example, taking the car scene as an example, the initial audio signal can be the navigation voice corresponding to the current road the car is traveling on, the favorite songs and music of the users in the car, etc.

[0039] S102. Determine a target filter coefficient corresponding to at least one target sound zone based on filter coefficients corresponding to a plurality of sound zones, wherein the filter coefficient is obtained by training a neural network model based on impulse response data received by an audio input device in the plurality of sound zones, and the target sound zone is one or more sound zones among the plurality of sound zones.

[0040] In order to overcome the limitations of audio output and provide users with a more personalized listening experience, the spatial area in the current scene is divided into multiple independent sound zones, including a first-category sound zone and a second-category sound zone. The first-category sound zone refers to a sound zone (bright zone) where audio is allowed to be received, that is, users in the first-category sound zone will hear the audio, and the second-category sound zone refers to a sound zone (dark zone) where audio is not allowed to be received, that is, users in the second-category sound zone will not hear the audio.

[0041] For example, still taking the car scenario as an example, if the initial audio signal is the navigation voice, the driver in the front seat of the car wants to hear the navigation voice, but the passenger in the back seat of the car does not want to hear the navigation voice. In this case, the front seat space area of ​​the car is the first type of sound zone, and the back seat space area of ​​the car is the second type of sound zone.

[0042] It is understandable that, for the same sound range, depending on the actual application situation, it may sometimes be the first type of sound range and sometimes be the second type of sound range.

[0043] The target sound zone is one or more sound zones among the multiple sound zones, and each target sound zone can be a first-category sound zone or a second-category sound zone. In order to enable users in the target sound zone to obtain a personalized listening experience, a neural network model is preliminarily trained based on impulse response data received by audio input devices in multiple sound zones to obtain corresponding multiple filter coefficients, wherein each filter coefficient corresponds to an audio output device and a corresponding sound zone. The target filter coefficient corresponding to each target sound zone is determined from the multiple filter coefficients. The audio input device includes but is not limited to a microphone.

[0044] In some embodiments, Figure 2 As shown, step S101 includes step S105 and step S106.

[0045] S105, collecting a plurality of impulse response data corresponding to a plurality of audio output devices to a plurality of audio input devices;

[0046] S106. Perform model training on the neural network model according to the plurality of impulse response data to obtain the trained neural network model.

[0047] Multiple audio output devices and multiple audio input devices are set in multiple independent sound zones, and impulse response data from each audio output device to each audio input device are collected respectively. Among them, Li represents the i-th audio output device, and Mi represents the i-th audio input device.

[0048] For example, taking the audio output device as a speaker and the audio input device as a microphone, the speaker plays the swept frequency signal in sequence, the microphone records it, and the impulse response data from the speaker to the microphone is obtained using a cross-correlation algorithm.

[0049] For example, considering that the scene environment may change, for example, in a car scene, the seat in the car may be adjusted, etc., it is not necessary to collect only the impulse response data of each audio output device to each audio input device at a time. As the input of the neural network model, the impulse response data from each audio output device to each audio input device is collected multiple times. As input to the neural network model.

[0050] Impulse response data As the training data for the neural network model, the neural network model is iteratively trained to obtain a trained neural network model. For each round of model training, the impulse response data Input the neural network model, output multiple undetermined filter coefficients, and record the multiple undetermined filter coefficients output by the trained neural network model, that is, the multiple undetermined filter coefficients output by the last round of model training, as filter coefficients Among them, Li represents the i-th audio output device, Zi represents the i-th audio zone information, and the audio zone information includes the physical space area information of the audio zone and the audio zone type information (first type audio zone or second type audio zone).

[0051] Exemplarily, the neural network model includes a feature layer, a fusion layer, and an output layer. In the model training stage, for example, Figure 3 As shown, the impulse response data As input, it is processed by the feature layer, fusion layer, and output layer respectively, and the model is trained to output the filter coefficients.

[0052] Among them, the feature layer is mainly composed of multi-layer convolutional layers, normalization layers, and activation layers; the fusion layer is composed of multi-layer convolutional layers, fully connected layers, and activation layers; the output layer is composed of multi-layer fully connected layers and activation layers.

[0053] Exemplarily, the feature layer includes multiple branch modules, and each branch module has the same structure. Considering that the final filter coefficient modulates the amplitude frequency to reduce the leakage of the bright area to the dark area, and uses the phase to offset, the feature layer can be set to include a first branch module and a second branch module. The first branch module and the second branch module have the same structure, and are both composed of multiple layers of convolutional layers, normalization layers, and activation layers. Using different branch modules to process and fuse the respective input data separately can improve the audio playback partition effect.

[0054] In some embodiments, Figure 4 As shown, step S106 may include sub-steps S1061 to S1064.

[0055] S1061, performing acoustic feature extraction on the impulse response data to obtain amplitude spectrum data and phase spectrum data corresponding to the impulse response data;

[0056] S1062, inputting the amplitude spectrum data and the phase spectrum data into different branch modules respectively to obtain corresponding first characteristic data and second characteristic data;

[0057] S1063, inputting the first feature data and the second feature data into the fusion layer for data fusion processing to obtain corresponding fused feature data;

[0058] S1064. Input the fused feature data into the output layer, and output a plurality of pending filter coefficients corresponding to each round of model training, wherein the filter coefficients are the pending filter coefficients output by the output layer in the last round of model training.

[0059] For example, the impulse response data Perform STFT (short-time Fourier transform) to extract acoustic features and obtain impulse response data Corresponding amplitude spectrum data and phase spectrum data, in each round of model training process, the amplitude spectrum data and phase spectrum data are respectively input into different branch modules of the feature layer in the neural network model to obtain the corresponding first feature data and second feature data. Figure 5 As shown, the amplitude spectrum data is input into the first branch module to obtain the corresponding first feature data, and the phase spectrum data is input into the second branch module to obtain the corresponding second feature data. Wherein, the first branch module and the second branch module are both composed of N layers of convolutional layers (Conv), normalization layers (Norm), and activation layers (Hardtanh).

[0060] Afterwards, the first feature data and the second feature data are fused through the fusion layer of the neural network model to obtain corresponding fused feature data, and the fused feature data is input into the output layer of the neural network model to output a plurality of pending filter coefficients corresponding to each round of model training until the model training is completed. The plurality of pending filter coefficients outputted from the last round of model training are the plurality of filter coefficients corresponding to each sound zone.

[0061] In some embodiments, during the neural network model training process, when a preset condition is met, it is determined that the neural network model training is completed; otherwise, if the preset condition is not met, the neural network model training continues until the preset condition is met and the model training process ends.

[0062] Exemplarily, satisfying the preset conditions includes:

[0063] The preset loss function corresponding to the neural network model converges; or

[0064] The function value of the preset loss function is less than or equal to a preset value; or

[0065] The number of iterative training times of the neural network model reaches a preset number threshold.

[0066] Among them, the preset value and preset number threshold can be flexibly set according to actual conditions, and are not specifically limited in this application.

[0067] In some embodiments, the preset loss function corresponding to the neural network model is related to the acoustic contrast level, the sound field matching degree, and the filter coefficient gain level; wherein the acoustic contrast level reflects the difference in sound intensity between the first type of sound zone and the second type of sound zone, the first type of sound zone is a sound zone that allows the reception of the initial audio signal, and the second type of sound zone is a sound zone that does not allow the reception of the initial audio signal; the sound field matching degree reflects the difference between the first frequency domain response data and the second frequency domain response data corresponding to the first type of sound zone, the first frequency domain response data is the impulse response data transmitted to the audio input device in the first type of sound zone, and the frequency domain response data corresponding to the frequency domain is obtained after filtering, and the second frequency domain response data is the corresponding expected frequency domain response data; the filter coefficient gain level reflects the overall gain of the multiple filter coefficients. The function value of the preset loss function can be determined based on the acoustic contrast level, the sound field matching degree, and the filter coefficient gain level.

[0068] In some embodiments, the audio signal processing method further includes:

[0069] The function value of the preset loss function is calculated according to the acoustic contrast level and the corresponding first preset adjustment factor, the sound field matching degree and the corresponding second preset adjustment factor, the filter coefficient gain level and the corresponding third preset adjustment factor.

[0070] Exemplarily, the function value of the preset loss function is calculated according to the following formula (1):

[0071] Loss=α*ACC_loss+β*PM_loss+γ*Gain_loss (1)

[0072] Among them, Loss is the function value of the preset loss function, ACC_loss is the sound contrast level, PM_loss is the sound field matching degree, Gain_loss is the filter coefficient gain level, α is the first preset adjustment factor, β is the second preset adjustment factor, γ is the third preset adjustment factor, α, β, γ can be any real number, which can be flexibly set according to actual conditions, and is not specifically limited in this application.

[0073] By using the preset loss function to solve the constraints on the sound contrast level, sound field matching, filter coefficient gain level, etc., the weights of each part can be adjusted more intuitively. It can be understood that more constraints can be flexibly added, and no specific restrictions are made in this application.

[0074] In some embodiments, the audio signal processing method further includes:

[0075] Summing each first frequency domain response data corresponding to the first type of sound zone to obtain a first sum value;

[0076] Summing each third frequency domain response data corresponding to the second type of sound zone to obtain a second sum value, wherein the third frequency domain response data is data obtained after filtering the frequency domain response data corresponding to the impulse response data transmitted to the audio input device in the second type of sound zone in the frequency domain;

[0077] The acoustic contrast level is determined according to the first sum value and the second sum value.

[0078] Exemplarily, the acoustic contrast level is calculated according to the following formula (2):

[0079]

[0080] Among them, ACC_loss indicates how many decibels the energy (sound intensity) of the first sound zone is higher than that of the second sound zone, B and D are the numbers of audio input devices in the first and second sound zones, Mi∈B and Mi∈D respectively indicate that the audio input devices belong to the first and second sound zones, Represents the transfer function from the i-th audio output device Li to the i-th audio input device Mi, that is, the impulse response data from the i-th audio output device Li to the i-th audio input device Mi The corresponding frequency domain response data in the frequency domain is It represents the undetermined filter coefficients corresponding to the audio zone from the i-th audio output device Li to the corresponding audio input device output during each round of training of the neural network model. express The frequency domain response data of the data to the position of the i-th audio input device Mi after filtering is described below for the convenience of distinguishing and describing. It is called the first frequency domain response data, and the position of the i-th audio input device Mi in the second type of sound zone corresponds to is called the third frequency domain response data, It is the superposition (first sum value) of the first frequency domain response data from all audio output devices to the positions of each audio input device in the first type of sound zone, that is, the signal class sum to each audio input device in the first type of sound zone when all audio output devices are working. It is the superposition (second sum) of the third frequency domain response data from all audio output devices to the positions of each audio input device in the second type of sound zone, that is, the signal class sum to each audio input device in the second type of sound zone when all audio output devices are working. a and b are preset coefficients respectively. For example, a can be set to 10 and b can be set to 20. Of course, a and b can also be set to other values, which are not specifically limited in this application.

[0081] In some embodiments, the audio signal processing method further includes:

[0082] Calculating the square value of the difference between each first frequency domain response data and the corresponding second frequency domain response data;

[0083] The sound field matching degree is determined according to each of the square values.

[0084] Exemplarily, the sound field matching degree is calculated according to the following formula (3):

[0085]

[0086] Among them, PM_loss is the sound field matching degree corresponding to the first type of sound zone, H des，Mi is the expected frequency domain response data corresponding to the position of the ith audio input device Mi in the first type of sound zone. For the convenience of distinguishing descriptions, it is referred to as the second frequency domain response data hereinafter. It should be noted that H des，Mi It can be obtained from the impulse response data collected in a certain sound effect scene, H des，MiIt can also be expressed as a standard Dirac function, or modulated by various impulse response data collected in different sound scenes. Considering the different requirements of users, H des，Mi Any modulation or synthesis can be performed without specific limitation.

[0087] In some embodiments, the audio signal processing method further includes:

[0088] determining a gain value for each of the filter coefficients;

[0089] The filter coefficient gain level is determined according to each of the gain values.

[0090] Exemplarily, the filter coefficient gain level is calculated according to the following formula (4):

[0091]

[0092] Where Gain_loss represents the filter coefficient gain level, L represents the number of all audio output devices, and gain represents the filter coefficient gain. When it is less than 0, gain is set to 0 and only counts The parts that are greater than 0. Of course, the constraint range of the filter coefficient gain can be adjusted according to actual conditions.

[0093] By iteratively training the neural network model, the function value Loss of the preset loss function is optimized to minimize the Loss, that is, the preset loss function converges and the model training is completed.

[0094] Of course, the model training may also be completed after the function value Loss of the preset loss function is less than or equal to the preset value. Alternatively, the model training is completed after the number of iterations of the neural network model training reaches a preset number threshold.

[0095] Then in the application stage, the various According to the audio output device that outputs the initial audio signal and at least one target audio zone, the target filter coefficient matched to each target audio zone can be determined.

[0096] S103: Filter the initial audio signal according to the target filter coefficient corresponding to at least one target sound zone to obtain at least one target audio signal.

[0097] Exemplarily, according to the target filter coefficient corresponding to each target sound zone, the initial audio signal is multiplied by the target filter coefficient corresponding to each target sound zone, and the initial audio signal is filtered to obtain the target audio signal corresponding to each target sound zone.

[0098] S104: Transmit each of the target audio signals to the corresponding target sound zone.

[0099] After obtaining the target audio signal corresponding to each target sound zone, each target audio signal is transmitted to the corresponding target sound zone through the audio output device. For example, taking the audio output device as a speaker, Figure 6 As shown, multiple Select the target filter coefficients corresponding to the loudspeaker and each target sound zone. The target audio signal is convolved with the initial audio signal and filtered to obtain the corresponding target audio signal, which is then transmitted from the speaker space to the control point (such as a microphone) in the corresponding target sound zone. Different target sound zones receive different target audio signals, thereby achieving the effect of audio zone playback.

[0100] By using impulse response data to drive the training of the neural network model, the corresponding filter coefficients are obtained, and then a fairly stable audio partition playback effect is obtained based on the corresponding filter coefficients, which has good stability. It is also applicable to complex scenes with multiple sound zones, and the audio partition playback effect is not affected.

[0101] Moreover, for changes in different scenarios, it is only necessary to collect different impulse response data or to achieve a fairly stable audio partition playback effect in various scenarios through a small amount of collection combined with simulation.

[0102] The audio signal processing method disclosed in the above embodiment, after obtaining the initial audio signal, determines the target filter coefficient corresponding to at least one target sound zone (the target sound zone is one or more sound zones among the multiple sound zones) based on the filter coefficients corresponding to multiple sound zones (the filter coefficients are obtained by training the neural network model based on the impulse response data received by the audio input devices in the multiple sound zones), and filters the initial audio signal according to the target filter coefficients corresponding to the at least one target sound zone to obtain at least one target audio signal, and then transmits each target audio signal to the corresponding target sound zone, thereby achieving the effect of audio zone playback and improving the intelligence of the audio output.

[0103] See also Figure 7 , Figure 7 is a schematic block diagram of a computer device provided in an embodiment of the present application. Figure 7 As shown, the computer device 300 includes one or more processors 301 and a memory 302. The processor 301 and the memory 302 are connected via a bus, such as an I2C (Inter-integrated Circuit) bus.

[0104] The one or more processors 301 work individually or collectively to execute the steps of the audio signal processing method provided in the above embodiment.

[0105] Specifically, the processor 301 may be a micro-controller unit (MCU), a central processing unit (CPU) or a digital signal processor (DSP).

[0106] Specifically, the memory 302 may be a Flash chip, a read-only memory (ROM) disk, an optical disk, a USB flash drive, or a mobile hard disk.

[0107] The processor 301 is used to run the computer program stored in the memory 302, and implement the steps of the audio signal processing method provided in the above embodiment when executing the computer program.

[0108] Exemplarily, the processor 301 is used to run the computer program stored in the memory 302, and implement the following steps when executing the computer program:

[0109] Obtaining an initial audio signal;

[0110] Determine a target filter coefficient corresponding to at least one target sound zone based on filter coefficients corresponding to the plurality of sound zones, wherein the filter coefficient is obtained by training a neural network model based on impulse response data received by an audio input device in the plurality of sound zones, and the target sound zone is one or more sound zones in the plurality of sound zones;

[0111] According to the target filter coefficient corresponding to at least one target sound zone, the initial audio signal is filtered to obtain at least one target audio signal;

[0112] Each of the target audio signals is transmitted to the corresponding target sound zone.

[0113] In some embodiments, before obtaining the initial audio signal, the processor 301 is configured to implement:

[0114] Collecting multiple impulse response data corresponding to multiple audio output devices to multiple audio input devices;

[0115] The neural network model is trained according to the plurality of impulse response data to obtain the trained neural network model.

[0116] In some embodiments, when a preset condition is met, it is determined that the training of the neural network model is completed, wherein meeting the preset condition includes:

[0117] The preset loss function corresponding to the neural network model converges; or

[0118] The function value of the preset loss function is less than or equal to a preset value; or

[0119] The number of iterative training times of the neural network model reaches a preset number threshold.

[0120] In some embodiments, the preset loss function is related to the sound contrast level, the sound field matching degree, and the filter coefficient gain level; the sound contrast level reflects the difference in sound intensity between the first category sound zone and the second category sound zone, the first category sound zone is a sound zone that is allowed to receive the initial audio signal, and the second category sound zone is a sound zone that is not allowed to receive the initial audio signal; the sound field matching degree reflects the difference between the first frequency domain response data and the second frequency domain response data corresponding to the first category sound zone, the first frequency domain response data is the impulse response data transmitted to the audio input device in the first category sound zone, and the frequency domain response data corresponding to the frequency domain is obtained after filtering, and the second frequency domain response data is the corresponding expected frequency domain response data; the filter coefficient gain level reflects the overall gain of the multiple filter coefficients.

[0121] In some embodiments, the processor 301 is further configured to implement:

[0122] The function value of the preset loss function is calculated according to the acoustic contrast level and the corresponding first preset adjustment factor, the sound field matching degree and the corresponding second preset adjustment factor, the filter coefficient gain level and the corresponding third preset adjustment factor.

[0123] In some embodiments, the processor 301 is further configured to implement:

[0124] Summing each first frequency domain response data corresponding to the first type of sound zone to obtain a first sum value;

[0125] Summing each third frequency domain response data corresponding to the second type of sound zone to obtain a second sum value, wherein the third frequency domain response data is data obtained after filtering the frequency domain response data corresponding to the impulse response data transmitted to the audio input device in the second type of sound zone in the frequency domain;

[0126] The acoustic contrast level is determined according to the first sum value and the second sum value.

[0127] In some embodiments, the processor 301 is further configured to implement:

[0128] Calculating the square value of the difference between each first frequency domain response data and the corresponding second frequency domain response data;

[0129] The sound field matching degree is determined according to each of the square values.

[0130] In some embodiments, the processor 301 is further configured to implement:

[0131] determining a gain value for each of the filter coefficients;

[0132] The filter coefficient gain level is determined according to each of the gain values.

[0133] In some embodiments, the neural network model includes a feature layer, a fusion layer, and an output layer, and the feature layer includes multiple branch modules. When the processor 301 implements the model training of the neural network model according to the multiple impulse response data, it is used to implement:

[0134] Extracting acoustic features from the impulse response data to obtain amplitude spectrum data and phase spectrum data corresponding to the impulse response data;

[0135] Inputting the amplitude spectrum data and the phase spectrum data into different branch modules respectively to obtain corresponding first characteristic data and second characteristic data;

[0136] Inputting the first feature data and the second feature data into the fusion layer for data fusion processing to obtain corresponding fused feature data;

[0137] The fused feature data is input into the output layer, and a plurality of undetermined filter coefficients corresponding to each round of model training are output, wherein the filter coefficients are the undetermined filter coefficients output by the output layer in the last round of model training.

[0138] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the steps of the audio signal processing method provided in the above embodiment.

[0139] The computer-readable storage medium may be an internal storage unit of the computer device described in any of the above embodiments, such as a hard disk or memory of the terminal device. The computer-readable storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the terminal device.

[0140] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An audio signal processing method, characterized in that: The method comprises: Obtaining an initial audio signal; Determine a target filter coefficient corresponding to at least one target sound zone based on filter coefficients corresponding to the plurality of sound zones, wherein the filter coefficient is obtained by training a neural network model based on impulse response data received by an audio input device in the plurality of sound zones, and the target sound zone is one or more sound zones in the plurality of sound zones; According to the target filter coefficient corresponding to at least one target sound zone, the initial audio signal is filtered to obtain at least one target audio signal; Each of the target audio signals is transmitted to the corresponding target sound zone.

2. The method according to claim 1, characterized in that Before obtaining the initial audio signal, the method includes: Collecting multiple impulse response data corresponding to multiple audio output devices to multiple audio input devices; The neural network model is trained according to the plurality of impulse response data to obtain the trained neural network model.

3. The method according to claim 2, characterized in that When a preset condition is met, determining that the neural network model training is completed, wherein meeting the preset condition includes: The preset loss function corresponding to the neural network model converges; or The function value of the preset loss function is less than or equal to a preset value; or The number of iterative training times of the neural network model reaches a preset number threshold.

4. The method according to claim 3, characterized in that The preset loss function is related to the sound contrast level, the sound field matching degree, and the filter coefficient gain level; the sound contrast level reflects the difference in sound intensity between the first category sound zone and the second category sound zone, the first category sound zone is a sound zone that is allowed to receive the initial audio signal, and the second category sound zone is a sound zone that is not allowed to receive the initial audio signal; the sound field matching degree reflects the difference between the first frequency domain response data and the second frequency domain response data corresponding to the first category sound zone, the first frequency domain response data is the impulse response data transmitted to the audio input device in the first category sound zone, and the frequency domain response data corresponding to the frequency domain is obtained after filtering, and the second frequency domain response data is the corresponding expected frequency domain response data; the filter coefficient gain level reflects the overall gain of the multiple filter coefficients.

5. The method according to claim 4, characterized in that The method further comprises: The function value of the preset loss function is calculated according to the acoustic contrast level and the corresponding first preset adjustment factor, the sound field matching degree and the corresponding second preset adjustment factor, the filter coefficient gain level and the corresponding third preset adjustment factor.

6. The method according to claim 4, characterized in that The method further comprises: Summing each first frequency domain response data corresponding to the first type of sound zone to obtain a first sum value; Summing each third frequency domain response data corresponding to the second type of sound zone to obtain a second sum value, wherein the third frequency domain response data is data obtained after filtering the frequency domain response data corresponding to the impulse response data transmitted to the audio input device in the second type of sound zone in the frequency domain; The acoustic contrast level is determined according to the first sum value and the second sum value.

7. The method according to claim 4, characterized in that The method further comprises: Calculating the square value of the difference between each first frequency domain response data and the corresponding second frequency domain response data; The sound field matching degree is determined according to each of the square values.

8. The method according to claim 4, characterized in that The method further comprises: determining a gain value for each of the filter coefficients; The filter coefficient gain level is determined according to each of the gain values.

9. The method according to claim 2, characterized in that: The neural network model includes a feature layer, a fusion layer, and an output layer, the feature layer includes a plurality of branch modules, and the neural network model is trained according to the plurality of impulse response data, including: Extracting acoustic features from the impulse response data to obtain amplitude spectrum data and phase spectrum data corresponding to the impulse response data; Inputting the amplitude spectrum data and the phase spectrum data into different branch modules respectively to obtain corresponding first characteristic data and second characteristic data; Inputting the first feature data and the second feature data into the fusion layer for data fusion processing to obtain corresponding fused feature data; The fused feature data is input into the output layer, and a plurality of undetermined filter coefficients corresponding to each round of model training are output, wherein the filter coefficients are the undetermined filter coefficients output by the output layer in the last round of model training.

10. A computer device, characterized in that: The computer device comprises: Memory and processor; Wherein, the memory is connected to the processor and is used to store programs; The processor is used to implement the steps of the audio signal processing method according to any one of claims 1 to 9 by running the program stored in the memory.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the steps of the audio signal processing method according to any one of claims 1 to 9.

Images