A signal processing method, computer device and storage medium

Abstract

The present application proposes a signal processing method, a computer device and a storage medium, the signal processing method comprising: obtaining an audio signal to be processed; calling an extended enhancement module of an audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain, and obtaining a target coding feature of the audio signal to be processed; calling an audio decoding module of an audio signal processing model to perform decoding processing on the target coding feature of the audio signal to be processed, and obtaining a target audio signal. Through the present application, the audio resolution of the audio signal to be processed can be improved.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to a signal processing method, a computer device, and a storage medium. Background Art

[0002] In today's era of continuous development of Internet technology, audio has become an indispensable part of people's daily life. In the field of audio signal processing, super-resolution is a method of enhancing sound quality. Its purpose is to use technical means such as signal processing to increase the resolution of audio, make up for the missing high-frequency details, and then improve the user's hearing experience.

[0003] At present, audio super-resolution (band extension) is mainly achieved by enhancing the sound quality based on deep neural networks (DNN) in the time domain and frequency domain of audio. Due to the complexity of audio time domain signals (especially music signals), it is difficult to learn the characteristics of super-resolution through neural networks. In the frequency domain, after performing short-time Fourier transform (STFT), only the amplitude is predicted and optimized, which makes it difficult to take into account phase information, which will have a negative impact on the sound quality. Therefore, this method is not effective enough in enhancing the sound quality. Summary of the invention

[0004] The embodiments of the present application propose a signal processing method, apparatus, computer device and storage medium, which can obtain target coding features based on the coding features of the audio signal to be processed, and then obtain the target audio signal based on the target coding features, thereby improving the audio resolution of the audio signal to be processed.

[0005] On the one hand, an embodiment of the present application provides a signal processing method, the method comprising:

[0006] Obtaining the audio signal to be processed;

[0007] Calling the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain the target coding features of the audio signal to be processed, wherein the feature enhancement processing is used to indicate that the time domain features of the audio signal to be processed are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the audio signal to be processed;

[0008] The audio decoding module of the audio signal processing model is called to decode the target coding features of the audio signal to be processed to obtain the target audio signal.

[0009] On the one hand, an embodiment of the present application provides a signal processing device, the device comprising:

[0010] An acquisition unit, used for acquiring an audio signal to be processed;

[0011] A processing unit, used to call the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain, so as to obtain the target coding features of the audio signal to be processed, wherein the feature enhancement processing is used to indicate that the time domain features of the audio signal to be processed are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the audio signal to be processed;

[0012] The processing unit is also used to call the audio decoding module of the audio signal processing model to decode the target coding features of the audio signal to be processed to obtain the target audio signal.

[0013] On the one hand, an embodiment of the present application provides a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the above-mentioned signal processing method.

[0014] On the one hand, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is read and executed by a processor of a computer device, the computer device executes the above-mentioned signal processing method.

[0015] On the one hand, the embodiments of the present application provide a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the above-mentioned signal processing method.

[0016] In the embodiment of the present application, when it is necessary to enhance the sound quality of the processed audio signal, the processed audio signal can be first obtained, and then the extended enhancement module of the audio signal processing model is called to perform feature enhancement processing on the coding features of the processed audio signal in the time domain, and the target coding features after the optimized features of the processed audio signal can be obtained. The feature enhancement processing is used to indicate that the time domain features of the processed audio signal are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the processed audio signal; Next, the target coding features of the processed audio signal are decoded by the audio decoding module of the audio signal processing model to obtain the target audio signal. It can be seen that in the embodiment of the present application, the neural network model (i.e., the audio signal processing model) can be used to directly learn the coding features of the processed audio signal in the time domain, and the optimized target coding features of the processed audio signal can be obtained. Then, based on the optimized target coding features, the target audio signal can be decoded. Since the target coding features are obtained after feature enhancement, the audio resolution of the target audio signal obtained based on the target coding features is also improved compared to the audio resolution of the processed audio signal. Therefore, the present application can improve the audio resolution of the processed audio signal. 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 It is a schematic diagram of the principle of an audio signal processing solution provided in an embodiment of the present application;

[0019] Figure 2 is a schematic diagram of the architecture of a signal processing system provided in an embodiment of the present application;

[0020] Figure 3 is a flowchart of a signal processing method provided in an embodiment of the present application;

[0021] Figure 4 It is a structural diagram of an audio encoding module of an audio signal processing model provided in an embodiment of the present application;

[0022] Figure 5 It is a structural diagram of an extended enhancement module of an audio signal processing model provided in an embodiment of the present application;

[0023] Figure 6 It is a structural diagram of an audio decoding module of an audio signal processing model provided in an embodiment of the present application;

[0024] Figure 7 It is a flowchart of a model training method provided in an embodiment of the present application;

[0025] Figure 8 is a structural schematic diagram of a signal processing device provided in an embodiment of the present application;

[0026] Fig. 9 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] 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 only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] It should be noted that the descriptions of "first", "second", etc. involved in the embodiments of the present application are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the technical features defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0029] The present application embodiment provides a signal processing solution that can be applied to the field of audio processing and achieve the effect of enhancing the sound quality of audio signals. Figure 1 , Figure 1It is a schematic diagram of the principle of an audio signal processing scheme provided by an embodiment of the present application. Among them, the signal processing scheme may specifically include: when it is necessary to enhance the sound quality of a section of an audio signal to be processed, the coding features of the audio signal to be processed can be obtained, wherein the method for determining the coding features of the audio signal to be processed can be determined based on a neural network encoder (encoder). Then, the extension enhancement module (Extension Module) of the audio signal processing model can be called to perform feature enhancement processing on the coding features of the audio signal to be processed about the time domain, so as to obtain the target coding features after the optimization of the audio signal to be processed. Finally, the target coding features of the audio signal to be processed are decoded by the audio decoding module of the audio signal processing model to obtain a target audio signal with a higher resolution. It can be seen that the signal processing method proposed in this scheme is to use the time domain feature to extract the audio encoder features (i.e., the coding features) as the input of the neural network to predict the encoder features of the high-resolution audio, avoiding the prediction difficulties caused by direct time domain waveform processing and the phase problems caused by short-time Fourier transform (STFT) frequency domain processing. Therefore, this application has a certain improvement on the generated high-resolution sound quality, avoiding some typical noise problems generated during frequency domain processing, thereby achieving the effect of improving audio resolution. In addition, it can effectively improve the height of the low-quality (too low spectrum, low time resolution) spectrum, and enhance the user's listening experience.

[0030] In order to better understand the embodiments of the present application, the professional terms involved in the embodiments of the present application are introduced below:

[0031] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0032] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, large-function testing technology, operation / interaction systems, mechatronics and other technologies. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0033] The so-called artificial intelligence cloud service is generally also called AIaaS (AIas a Service, Chinese for "AI as a Service"). This is the current mainstream service mode of artificial intelligence platforms. Specifically, the AIaaS platform will split several common AI services and provide independent or packaged services in the cloud. This service model is similar to opening an AI theme mall: all developers can access and use one or more artificial intelligence services provided by the platform through API interfaces. Some senior developers can also use the AI ​​framework and AI infrastructure provided by the platform to deploy and operate their own cloud artificial intelligence services.

[0034] The signal processing method involved in this application can be encapsulated as an artificial intelligence service, and only one interface is exposed to the outside. When the version function test of the product or application software involved in this application is required in a certain business scenario, by calling this interface, the audio signal to be processed can be obtained and the audio signal processing model in the embodiment of this application can be used to process the audio signal to be processed to generate a super-resolution target audio signal.

[0035] In one possible implementation, the signal processing solution of the present application can be combined with blockchain technology. For example, the acquired audio signal to be processed, the generated target audio signal and other data can be uploaded to the blockchain for storage, which can ensure that these data on the blockchain are not easily tampered with.

[0036] Please refer to Figure 2 , Figure 2 : is a schematic diagram of the architecture of a signal processing system provided in an embodiment of the present application. The architecture diagram of the signal processing system includes: a server 240 and a terminal device cluster, wherein the terminal device cluster may include: a terminal device 210, a terminal device 220, a terminal device 230, etc. The terminal device cluster and the server 240 may be directly or indirectly connected via wired or wireless communication, and the present application does not limit this.

[0037] Figure 1The server 240 shown can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0038] Figure 1 The terminal device 210, terminal device 220, terminal device 230, etc. shown in the figure can be a mobile phone, a tablet computer, a laptop computer, a PDA, a mobile Internet device (MID), a vehicle, a vehicle-mounted device, a roadside device, an aircraft, a wearable device, such as a smart watch, a smart bracelet, a pedometer, etc., and other smart devices with signal processing functions.

[0039] In a possible implementation, taking the terminal device 210 as an example, first, the audio signal to be processed is obtained, and then the terminal device 210 can send the audio signal to be processed to the server 240. The server 240 can first obtain the coding features of the audio signal to be processed, and then the server 240 calls the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed about the time domain, and obtains the target coding features of the audio signal to be processed. Next, the server 240 calls the audio decoding module of the audio signal processing model to decode the target coding features of the audio signal to be processed, and obtains the target audio signal. Finally, the server 240 can send the generated target audio signal to the terminal device 210.

[0040] Of course, the operations of performing feature enhancement processing on the coding features of the audio signal to be processed in the time domain and decoding the target coding features of the audio signal to be processed are not necessarily performed by the server 240, but may also be performed by the terminal device 210. In addition, determining the coding features of the audio signal to be processed is not necessarily performed by the server 240, but may also be performed by the terminal device 210, and this embodiment of the present application does not specifically limit this.

[0041] In a possible implementation, the signal processing system provided in the embodiment of the present application can be deployed on a node of the blockchain. For example, the server 240 and each terminal device included in the terminal device cluster can be regarded as a node device of the blockchain to form a blockchain network. Therefore, the signal processing process for the audio signal to be processed in the present application can be executed on the blockchain, which can not only ensure the fairness and justice of the signal processing process, but also make the signal processing process traceable and improve the security of the signal processing process.

[0042] It can be understood that the system architecture diagram described in the embodiment of the present application is for more clearly illustrating the technical solution of the embodiment of the present application, and does not constitute a limitation on the technical solution provided in the embodiment of the present application. A person of ordinary skill in the art can know that with the evolution of the system architecture and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is also applicable to similar technical problems.

[0043] Based on the above analysis, the following Figure 3 The signal processing method of this application is described. Figure 3 , Figure 3 1 is a flow chart of a signal processing method provided by an embodiment of the present application. The signal processing method can be executed by a computer device, which can be a terminal device or a server. The signal processing method can also be executed by the terminal device and the server together. Figure 3 As shown, the signal processing method may include steps S310 to S330. Among them:

[0044] S310: Acquire an audio signal to be processed.

[0045] In an embodiment of the present application, in a possible implementation, the audio signal to be processed can be obtained online in real time, for example, downloaded locally from an audio playback software (such as XX music software) in real time, thereby obtaining the corresponding audio signal to be processed. In addition, after the audio signal is downloaded locally, the audio processing software can be used to perform audio enhancement processing on the audio signal downloaded locally, and then the audio signal after the audio enhancement processing is used as the audio signal to be processed in the present application. Among them, the audio enhancement processing may include but is not limited to: speed change, pitch change, increase or decrease the volume, and add attack to the original audio, such as adding applause, other off-site sounds, etc.

[0046] Of course, in another possible implementation, the audio signal to be processed acquired by the computer device may be input by the user. For example, the user inputs a certain specified audio signal into the computer device, and the computer device uses the specified audio signal input by the user as the audio signal to be processed.

[0047] It should be noted that the audio resolution of the audio signal to be processed in the embodiment of the present application is relatively low, that is, the audio signal to be processed is an audio signal that requires sound quality enhancement processing. In the present application, the audio signal to be processed may also be referred to as a low-resolution signal, and the audio signal to be processed may be represented as an LR waveform.

[0048] S320: Call the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the time domain coding features of the audio signal to be processed to obtain the target coding features of the audio signal to be processed. The feature enhancement processing is used to indicate that the time domain features of the audio signal to be processed are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the audio signal to be processed.

[0049] Among them, the coding features of the audio signal to be processed are the time domain features of the audio signal to be processed. The extended enhancement module in this scheme is used to enhance the time domain features of the audio signal to be processed, so that the lower resolution coding features of the audio signal to be processed can be enhanced to obtain higher resolution target coding features.

[0050] In the embodiment of the present application, the audio signal processing model can be a neural network model of any structure composed of a coding module, an extended enhancement module, and a decoding module. The coding module is generally composed of 1-DCNN (one-dimensional convolution); the decoding module is generally composed of 1-DCNN convolution; the extended enhancement module can be composed of a neural network model of any structure, and the embodiment of the present application does not specifically limit the network structure of the extended enhancement module. For example, the audio signal processing model can be specifically an RNN (Recurrent Neural Network, RNN) model, LSTM (Long Short-Term Memory Network, LSTM, LongShort-Term Memory), GRU (Gated Recurrent Neural Network, Gated Recurrent Neural Network) model, CNN (Convolutional Neural Network, Convolutional Neural Network), or the neural network model can also be other types of network models, such as recursive neural network models, etc., and the present application does not limit this.

[0051] In one possible implementation, before calling the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed with respect to the time domain, the audio encoding module of the audio signal processing model can be called to perform encoding processing on the audio signal to be processed, thereby obtaining the coding features of the audio signal to be processed.

[0052] In a possible implementation, the audio encoding module of the audio signal processing model includes M one-dimensional convolution kernels, a regularization function, and an activation function, where M is a positive integer. The process in which the computer device calls the audio encoding module of the audio signal processing model to encode the audio signal to be processed and obtain the encoding features of the audio signal to be processed may include: first, using M one-dimensional convolution kernels to perform convolution processing on the audio signal to be processed to obtain a convolution feature vector corresponding to the audio signal to be processed. Then, using the regularization function and the activation function, the convolution feature vector is processed to obtain the encoding features of the audio signal to be processed.

[0053] For example, see Figure 4 , Figure 4 Schematic diagram of the structure of an audio coding module of an audio signal processing model provided in an embodiment of the present application. Figure 4 It can be seen that the audio encoding module of the audio signal processing model can be composed of several (represented as M) one-dimensional convolutions (1-Dimension, 1-D). The audio signal to be processed can be input into the audio encoding module, and then the audio signal to be processed is convolved through M 1-D convolution kernels to obtain the convolution feature vector corresponding to the audio signal to be processed. Then, the convolution feature vector corresponding to the audio signal to be processed is normalized by using a regularization function to obtain a normalized convolution feature vector. In addition, the normalized convolution feature vector can be activated by using an activation function to determine the coding features of the audio signal to be processed. Among them, the regularization function can include but is not limited to: batch norm function, etc., and the activation function can include but is not limited to: Sigmoid function, tanh function, Relu function, etc.

[0054] Further, the following is a detailed description of how to use a one-dimensional convolution kernel to perform convolution processing on the audio signal to be processed. In this application, the audio signal to be processed can be represented as: x_LR(n), n = 0, 1, ..., N. First, the time domain audio signal to be processed is divided into 50% overlapping segments (segment length is L), and T segmented segments are obtained, denoted as x_LR(k)∈R T×L After several (1-D convolution CNN+batch norm+ReLU) steps, the encoder feature is obtained, denoted as Enc_LR∈R T×K .

[0055] In summary, by encoding the audio signal to be processed using the audio encoding module of the audio signal processing model, the low resolution (LR) encoder feature (encoding feature) of the audio signal to be processed can be determined. The LR encoder feature of the audio signal to be processed can be expressed as: Enc_LR∈RT×K .

[0056] In a possible implementation, the coding features of the audio signal to be processed obtained based on the audio coding module may include a first unit coding vector and a second unit coding vector. Then, the process in which the computer device calls the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed about the time domain to obtain the target coding features of the audio signal to be processed may include: first, based on the extended enhancement module of the audio signal processing model, the first unit coding vector and the initial hidden vector, determine the first hidden vector and the first output vector of the first unit coding vector. Then, based on the extended enhancement module of the audio signal processing model, the second unit coding vector and the first hidden vector, determine the second hidden vector and the second output vector of the second unit coding vector. Finally, the first output vector and the second output vector may be fully connected to obtain the target coding features of the audio signal to be processed.

[0057] In specific implementation, the extension network in the embodiment of the present application can be a neural network of any structure, for example, the neural network includes but is not limited to: the neural networks of the above-mentioned LSTM, CNN, DNN and other structures, and the embodiment of the present application does not make specific limitations on this.

[0058] For example, take the extended enhancement module as an LSTM network structure, see Figure 5 , Figure 5 Schematic diagram of the structure of an extended enhancement module of an audio signal processing model provided in an embodiment of the present application. Figure 5 It can be seen that the encoding feature Enc_LR∈R of the audio signal to be processed can be T×K Input to Figure 5 The extended enhancement module shown in the figure performs feature enhancement learning, and then through the learning of the extended enhancement module, HR (High Resolution) encoder features (encoding features) can be generated. Figure 5 The x1 and x2 shown can be Enc_LR∈R T×K The first unit encoding vector and the second unit encoding vector included in the , h1 can be the initial hidden vector, h2 can be the first hidden vector, wherein the first unit encoding vector x1 can refer to the encoding feature Enc_LR∈R T×K Any vector (any row vector or any column vector) in (matrix) and the second unit encoding vector x2 can refer to the encoding feature Enc_LR∈R T×K Any vector (any row vector or any column vector) except x1 in . It should be noted that, for example, if the first unit encoding vector x1 is the encoding feature Enc_LR∈RT×K Any row vector in, then the second unit encoding vector x2 is the encoding feature Enc_LR∈R T×K Any row vector except x1 in ; For example, if the first unit encoding vector x1 is the encoding feature Enc_LR∈R T×K Any column vector in, then the second unit encoding vector x2 is the encoding feature Enc_LR∈R T×K Any column vector except x1 in Enc_LR∈R. Similarly, x3 can be Enc_LR∈R T×K The third unit coding vector included in , and h3 can be the second hidden vector. Similarly, the second unit coding vector x3 and the second hidden vector h3 are input into the extended enhancement module of the audio signal processing model to determine the second hidden vector and the second output vector of the second unit coding vector. Of course, Figure 5 The structure of the extended enhancement module in the embodiment of the present application is shown as a schematic diagram, and the structure of the extended enhancement module in the present application is not specifically limited. Figure 5 The extended enhancement module shown in the figure adds an attention module (attention module) or a self-attention module (self-attention module) to better Enc_LR∈R T×K Features are learned to obtain high-resolution coding features. The high-resolution coding features (i.e., target audio features) can be expressed as: Enc_HR∈R T×K .

[0059] In this way, the neural network model can be used to directly process the LR encoder features (i.e., low-resolution encoding features) of the audio signal to be processed, and predict the HR encoder features (i.e., high-resolution encoding features) of the audio signal to be processed.

[0060] S330: Calling the audio decoding module of the audio signal processing model to decode the target coding features of the audio signal to be processed to obtain a target audio signal.

[0061] In the embodiment of the present application, the network structure of the audio decoding module may be the same as or different from the network structure of the audio encoding module.

[0062] In a possible implementation, the audio decoding module of the audio signal processing model may include N one-dimensional convolution kernels, a regularization function, and an activation function, where N is a positive integer. The process in which a computer device calls the audio decoding module of the audio signal processing model to decode the target coding features of the audio signal to be processed and obtain the target audio signal may include: first, using N one-dimensional convolution kernels to perform convolution processing on the target coding features of the audio signal to be processed to obtain a convolution feature vector corresponding to the target coding features. Then, using the regularization function and the activation function, the convolution feature vector corresponding to the target coding features is fully connected to obtain the target audio signal.

[0063] For example, see Figure 6 , Figure 6 Schematic diagram of the structure of an audio decoding module of an audio signal processing model provided in an embodiment of the present application. Figure 6 It can be seen that the audio decoding module of the audio signal processing model can be composed of several (which can be expressed as N) 1-D convolutions. Enc_HR∈R T×K Input to the audio decoding module, and then processed by N 1-D convolutions to obtain Enc_HR∈R T×K The corresponding convolution feature vector x_HR(k)∈R T×L The obtained segment signal x_HR(k)∈ T×L The high-resolution target audio signal (HR waveform) is predicted by 50% overlap splicing. The HR waveform can be expressed as: x_HR(n)∈R 1×N . Thus, the high-resolution target audio signal (HR waveform) is predicted. The HR waveform can be expressed as: x_HR(n)∈R 1×N2 .

[0064] It should be noted that in the embodiment of the present application, the detailed steps of "using a one-dimensional convolution kernel to perform convolution processing on the target coding features of the audio signal to be processed to obtain a convolution feature vector corresponding to the target coding features" can be specifically referred to the detailed steps of "using a one-dimensional convolution kernel to perform convolution processing on the audio signal to be processed to obtain a convolution feature vector corresponding to the audio signal to be processed" involved in the above step S320, and the embodiment of the present application is not specifically limited here.

[0065] In the embodiment of the present application, when it is necessary to enhance the sound quality of the processed audio signal, the processed audio signal can be first obtained, and then the extended enhancement module of the audio signal processing model is called to perform feature enhancement processing on the coding features of the processed audio signal in the time domain, and the target coding features after the optimized features of the processed audio signal can be obtained. The feature enhancement processing is used to indicate that the time domain features of the processed audio signal are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the processed audio signal; Next, the target coding features of the processed audio signal are decoded by the audio decoding module of the audio signal processing model to obtain the target audio signal. It can be seen that in the embodiment of the present application, the neural network model (i.e., the audio signal processing model) can be used to directly learn the coding features of the processed audio signal in the time domain, and the optimized target coding features of the processed audio signal can be obtained. Then, based on the optimized target coding features, the target audio signal can be decoded. Since the target coding features are obtained after feature enhancement, the audio resolution of the target audio signal obtained based on the target coding features is also improved compared to the audio resolution of the processed audio signal. Therefore, the present application can improve the audio resolution of the processed audio signal.

[0066] Based on the above analysis, the following is a detailed description of how to train the audio signal processing model. Figure 7 , Figure 7 1 is a flow chart of a model training method provided in an embodiment of the present application. The model training method can be executed by a computer device, which can be a terminal device or a server. The signal processing method can also be executed by a terminal device and a server together. Figure 7 As shown, the model training method may include steps S710 to S730. Among them:

[0067] S710: Acquire a sample audio set, where the sample audio set includes a plurality of sample audio signals and a reference audio signal corresponding to each sample audio signal, and a resolution of the reference audio signal is greater than a resolution of the sample audio signal.

[0068] In an embodiment of the present application, each sample audio signal in the sample audio set can be obtained online in real time, for example, downloaded locally in real time from an audio playback software (such as XX music software), so as to obtain the corresponding sample audio signal. At the same time, a reference audio signal corresponding to each sample audio signal can also be obtained. Among them, the sample audio signal can refer to an audio signal with a lower resolution, or an audio signal doped with other noises such as noise and reverberation; and the corresponding reference audio signal can be an audio signal after preprocessing the sample audio signal, and the preprocessing can be, for example, filtering processing, vocal separation processing, noise reduction processing, etc. Then, it can be considered that the resolution of the reference audio signal is greater than the resolution of the sample audio signal, or the sample audio signal is a low-resolution audio signal, and the reference audio signal is a high-resolution audio signal.

[0069] S720: Calling the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal for each sample audio signal.

[0070] In a possible implementation, a process in which a computer device calls an initial neural network model to perform signal processing on sample audio signals in a sample audio set to obtain a predicted audio signal for each sample audio signal may include: first, the computer device calls an audio encoding module of the initial neural network model to encode the sample audio signals in the sample audio set to obtain a first encoding feature of the sample audio signal. Then, the computer device calls an extended enhancement module of the initial neural network model to perform feature enhancement processing on the first encoding feature of the sample audio signal to obtain a second encoding feature of the sample audio signal. Finally, the computer device calls an audio decoding module of the initial neural network model to decode the second encoding feature of the sample audio signal to obtain a predicted audio signal of the sample audio signal.

[0071] In a specific implementation, the audio encoding module of the initial neural network model may include M one-dimensional convolution kernels, a regularization function, and an activation function, where M is a positive integer. The process in which a computer device calls the audio encoding module of the initial neural network model to encode a sample audio signal in a sample audio set to obtain a first encoding feature of the sample audio signal may include: first, convolution processing is performed on the sample audio signal using M one-dimensional convolution kernels to obtain a convolution feature vector corresponding to the sample audio signal. Then, the convolution feature vector corresponding to the sample audio signal is fully connected using a regularization function and an activation function to obtain a first encoding feature of the sample audio signal.

[0072] In addition, the first coding feature of the sample audio signal may include a first unit coding vector and a second unit coding vector. The process in which the computer device calls the extended enhancement module of the initial neural network model to perform feature enhancement processing on the first coding feature of the sample audio signal to obtain the second coding feature of the sample audio signal may include: first, based on the extended enhancement module of the initial neural network model, the first unit coding vector and the initial hidden vector, determine the first hidden vector and the first output vector of the first unit coding vector. Then, based on the extended enhancement module of the initial neural network model, the second unit coding vector and the first hidden vector, determine the second hidden vector and the second output vector of the second unit coding vector. Finally, the first output vector and the second output vector are fully connected to obtain the second coding feature of the audio signal to be processed.

[0073] Next, the audio decoding module of the initial neural network model may include N one-dimensional convolution kernels, a regularization function, and an activation function, where N is a positive integer. The process in which the computer device calls the audio decoding module of the initial neural network model to decode the second coding feature of the sample audio signal to obtain a predicted audio signal of the sample audio signal may include: first, using N one-dimensional convolution kernels to perform convolution processing on the second coding feature of the sample audio signal to obtain a convolution feature vector corresponding to the second coding feature. Then, using the regularization function and the activation function, the convolution feature vector corresponding to the second coding feature is fully connected to obtain a predicted audio signal of the sample audio signal.

[0074] S730: Train an initial neural network model according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain an audio signal processing model.

[0075] In one possible implementation, the computer device adjusts the model parameters of the audio encoding module of the initial neural network model, the extended enhancement module of the initial neural network model, and the audio decoding module of the initial neural network model according to the difference between the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal. Then, when the adjusted audio encoding module of the initial neural network model, the extended enhancement module of the initial neural network model, and the audio decoding module of the initial neural network model all meet the model convergence conditions, the initial neural network model with the adjusted model parameters is used as the audio signal processing model. The difference between the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal may include: the difference between the predicted audio signal and the reference audio signal in the resolution dimension.

[0076] Among them, the so-called model convergence condition may refer to: when the number of training times of the initial neural network model reaches a preset training threshold, for example, 100 times, the initial neural network model can be considered to meet the model convergence condition, that is, the initial neural network model after training 100 times is used as the audio signal processing model. When the difference data between the resolution of the predicted audio signal predicted by the model and the reference audio signal is less than the error threshold, the initial neural network model can be considered to meet the model convergence condition. When the change between the predicted audio signals obtained by two consecutive trainings of the initial neural network model is less than the change threshold, the neural network model can be considered to meet the model convergence condition.

[0077] In addition, when training the initial neural network model, the three modules of audio encoding module, extension enhancement module and audio decoding module are jointly optimized, and the goal of optimizing the initial neural network model is Among them, x_HR(n) refers to the reference audio signal of each sample audio signal, It refers to the predicted audio signal of each sample audio signal.

[0078] Through the embodiment of the present application, the initial neural network model composed of three modules, namely, the audio encoding module, the extended enhancement module and the audio decoding module, is trained by using the low-resolution sample audio signals in the sample audio set and the corresponding high-resolution reference audio signals to obtain an audio signal processing model. Since the encoder coding features in the embodiment of the present application are completely learned by the neural network model, they are more adaptable to super-resolution scenarios, and also avoid the phase problem caused by the audio signal after frequency domain processing. Therefore, the audio signal processing model trained by the embodiment of the present application can improve the resolution of the audio signal.

[0079] See also Figure 8 , Figure 8 800 is a schematic diagram of a signal processing device provided in an embodiment of the present application. Figure 2 Computer device in the corresponding method embodiment. The signal processing device 800 may be a computer program (including program code) running in a computer device, for example, the signal processing device 800 is an application software; the device may be used to execute the corresponding steps in the method provided in the embodiment of the present application. The signal processing device 800 may include:

[0080] An acquisition unit 801 is used to acquire an audio signal to be processed;

[0081] The processing unit 802 is used to call the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain the target coding features of the audio signal to be processed, wherein the feature enhancement processing is used to indicate that the time domain features of the audio signal to be processed are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the audio signal to be processed;

[0082] The processing unit 802 is further configured to call the audio decoding module of the audio signal processing model to perform decoding processing on the target coding features of the audio signal to be processed, so as to obtain a target audio signal.

[0083] In a possible implementation, the processing unit 802 calls the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain, and before obtaining the target coding features of the audio signal to be processed, is further used to perform the following operations:

[0084] The audio encoding module of the audio signal processing model is called to encode the audio signal to be processed to obtain the encoding features of the audio signal to be processed.

[0085] In a possible implementation, the audio encoding module of the audio signal processing model includes M one-dimensional convolution kernels, a regularization function, and an activation function, where M is a positive integer;

[0086] When the processing unit 802 calls the audio encoding module of the audio signal processing model to encode the audio signal to be processed and obtains the encoding features of the audio signal to be processed, it is used to perform the following operations:

[0087] Using M one-dimensional convolution kernels to perform convolution processing on the audio signal to be processed, to obtain a convolution feature vector corresponding to the audio signal to be processed;

[0088] The convolution feature vector is processed using the regularization function and the activation function to obtain the coding features of the audio signal to be processed.

[0089] In a possible implementation, the coding feature of the audio signal to be processed includes a first unit coding vector and a second unit coding vector;

[0090] When the processing unit 802 calls the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain the target coding features of the audio signal to be processed, it is used to perform the following operations:

[0091] Inputting the first unit code vector and the initial hidden vector into an extended enhancement module of an audio signal processing model to determine a first hidden vector and a first output vector of the first unit code vector;

[0092] Inputting the second unit code vector and the first hidden vector into an extended enhancement module of an audio signal processing model to determine a second hidden vector and a second output vector of the second unit code vector;

[0093] The first output vector and the second output vector are fully connected to obtain target coding features of the audio signal to be processed.

[0094] In a possible implementation, the audio decoding module of the audio signal processing model includes N one-dimensional convolution kernels, where N is a positive integer;

[0095] When the processing unit 802 calls the audio decoding module of the audio signal processing model to decode the target coding features of the audio signal to be processed and obtains the target audio signal, it is used to perform the following operations:

[0096] The target coding features of the audio signal to be processed are convolved using N one-dimensional convolution kernels to obtain the target audio signal.

[0097] In a possible implementation, the processing unit 802 is further configured to perform the following operations:

[0098] Acquire a sample audio set, where the sample audio set includes a plurality of sample audio signals and a reference audio signal corresponding to each sample audio signal, and a resolution of the reference audio signal is greater than a resolution of the sample audio signal;

[0099] Calling the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal for each sample audio signal;

[0100] An initial neural network model is trained according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain an audio signal processing model.

[0101] In a possible implementation, when the processing unit 802 calls the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal for each sample audio signal, the processing unit 802 is configured to perform the following operations:

[0102] Calling the audio encoding module of the initial neural network model to encode the sample audio signal in the sample audio set to obtain a first encoding feature of the sample audio signal;

[0103] Calling the extended enhancement module of the initial neural network model to perform feature enhancement processing on the first encoding feature of the sample audio signal to obtain a second encoding feature of the sample audio signal;

[0104] The audio decoding module of the initial neural network model is called to decode the second encoded feature of the sample audio signal to obtain a predicted audio signal of the sample audio signal.

[0105] In a possible implementation, the processing unit 802 is configured to perform the following operations when training the initial neural network model according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain the audio signal processing model:

[0106] adjusting, according to a difference between a predicted audio signal for each sample audio signal and a reference audio signal for each sample audio signal, respectively, an audio encoding module of the initial neural network model, an audio encoding module of the initial neural network model, and a model parameter of the audio encoding module of the initial neural network model;

[0107] When the adjusted initial neural network model meets the model convergence condition, the initial neural network model after the model parameters are adjusted is used as the audio signal processing model.

[0108] In the embodiment of the present application, when it is necessary to enhance the sound quality of the processed audio signal, the processed audio signal can be first obtained, and then the extended enhancement module of the audio signal processing model is called to perform feature enhancement processing on the coding features of the processed audio signal in the time domain, and the target coding features after the optimized features of the processed audio signal can be obtained. The feature enhancement processing is used to indicate that the time domain features of the processed audio signal are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the processed audio signal; Next, the target coding features of the processed audio signal are decoded by the audio decoding module of the audio signal processing model to obtain the target audio signal. It can be seen that in the embodiment of the present application, the neural network model (i.e., the audio signal processing model) can be used to directly learn the coding features of the processed audio signal in the time domain, and the optimized target coding features of the processed audio signal can be obtained. Then, based on the optimized target coding features, the target audio signal can be decoded. Since the target coding features are obtained after feature enhancement, the audio resolution of the target audio signal obtained based on the target coding features is also improved compared to the audio resolution of the processed audio signal. Therefore, the present application can improve the audio resolution of the processed audio signal.

[0109] See also Fig. 9 , Fig. 9 900 is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. The computer device 900 is used to execute Figure 3The steps performed by the computer device in the corresponding method embodiment are as follows: the computer device 900 includes: one or more processors 910; one or more input devices 920, one or more output devices 930 and a memory 940. The processor 910, the input device 920, the output device 930 and the memory 940 are connected via a bus 950. The memory 940 is used to store a computer program, and the computer program includes program instructions. The processor 910 is used to call the program instructions stored in the memory 940 to perform the following operations:

[0110] Obtaining the audio signal to be processed;

[0111] Calling the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain the target coding features of the audio signal to be processed, wherein the feature enhancement processing is used to indicate that the time domain features of the audio signal to be processed are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the audio signal to be processed;

[0112] The audio decoding module of the audio signal processing model is called to decode the target coding features of the audio signal to be processed to obtain the target audio signal.

[0113] In a possible implementation, the processor 910 calls the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain, and before obtaining the target coding features of the audio signal to be processed, is further used to perform the following operations:

[0114] The audio encoding module of the audio signal processing model is called to encode the audio signal to be processed to obtain the encoding features of the audio signal to be processed.

[0115] In a possible implementation, the audio encoding module of the audio signal processing model includes M one-dimensional convolution kernels, a regularization function, and an activation function, where M is a positive integer;

[0116] When the processor 910 calls the audio encoding module of the audio signal processing model to encode the audio signal to be processed and obtains the encoding features of the audio signal to be processed, it is used to perform the following operations:

[0117] Using M one-dimensional convolution kernels to perform convolution processing on the audio signal to be processed, to obtain a convolution feature vector corresponding to the audio signal to be processed;

[0118] The convolution feature vector is processed using the regularization function and the activation function to obtain the coding features of the audio signal to be processed.

[0119] In a possible implementation, the coding feature of the audio signal to be processed includes a first unit coding vector and a second unit coding vector;

[0120] When the processor 910 calls the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain the target coding features of the audio signal to be processed, it is used to perform the following operations:

[0121] Inputting the first unit code vector and the initial hidden vector into an extended enhancement module of an audio signal processing model to determine a first hidden vector and a first output vector of the first unit code vector;

[0122] Inputting the second unit code vector and the first hidden vector into an extended enhancement module of an audio signal processing model to determine a second hidden vector and a second output vector of the second unit code vector;

[0123] The first output vector and the second output vector are fully connected to obtain target coding features of the audio signal to be processed.

[0124] In a possible implementation, the audio decoding module of the audio signal processing model includes N one-dimensional convolution kernels, where N is a positive integer;

[0125] When the processor 910 calls the audio decoding module of the audio signal processing model to decode the target coding feature of the audio signal to be processed and obtains the target audio signal, it is used to perform the following operations:

[0126] The target coding features of the audio signal to be processed are convolved using N one-dimensional convolution kernels to obtain the target audio signal.

[0127] In a possible implementation, the processor 910 is further configured to perform the following operations:

[0128] Acquire a sample audio set, where the sample audio set includes a plurality of sample audio signals and a reference audio signal corresponding to each sample audio signal, and a resolution of the reference audio signal is greater than a resolution of the sample audio signal;

[0129] Calling the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal for each sample audio signal;

[0130] An initial neural network model is trained according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain an audio signal processing model.

[0131] In a possible implementation, when the processor 910 calls the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal for each sample audio signal, the processor 910 is configured to perform the following operations:

[0132] Calling the audio encoding module of the initial neural network model to encode the sample audio signal in the sample audio set to obtain a first encoding feature of the sample audio signal;

[0133] Calling the extended enhancement module of the initial neural network model to perform feature enhancement processing on the first encoding feature of the sample audio signal to obtain a second encoding feature of the sample audio signal;

[0134] The audio decoding module of the initial neural network model is called to decode the second encoded feature of the sample audio signal to obtain a predicted audio signal of the sample audio signal.

[0135] In a possible implementation, the processor 910 is configured to perform the following operations when training the initial neural network model according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain the audio signal processing model:

[0136] adjusting, according to a difference between a predicted audio signal for each sample audio signal and a reference audio signal for each sample audio signal, respectively, an audio encoding module of the initial neural network model, an audio encoding module of the initial neural network model, and a model parameter of the audio encoding module of the initial neural network model;

[0137] When the adjusted initial neural network model meets the model convergence condition, the initial neural network model after the model parameters are adjusted is used as the audio signal processing model.

[0138] In the embodiment of the present application, when it is necessary to enhance the sound quality of the processed audio signal, the processed audio signal can be first obtained, and then the extended enhancement module of the audio signal processing model is called to perform feature enhancement processing on the coding features of the processed audio signal in the time domain, and the target coding features after the optimized features of the processed audio signal can be obtained. The feature enhancement processing is used to indicate that the time domain features of the processed audio signal are enhanced, and the resolution of the target coding features is higher than the resolution of the coding features of the processed audio signal; Next, the target coding features of the processed audio signal are decoded by the audio decoding module of the audio signal processing model to obtain the target audio signal. It can be seen that in the embodiment of the present application, the neural network model (i.e., the audio signal processing model) can be used to directly learn the coding features of the processed audio signal in the time domain, and the optimized target coding features of the processed audio signal can be obtained. Then, based on the optimized target coding features, the target audio signal can be decoded. Since the target coding features are obtained after feature enhancement, the audio resolution of the target audio signal obtained based on the target coding features is also improved compared to the audio resolution of the processed audio signal. Therefore, the present application can improve the audio resolution of the processed audio signal.

[0139] In addition, it should be pointed out here that: the embodiment of the present application also provides a computer storage medium, and the computer storage medium stores a computer program, and the computer program includes program instructions. When the processor executes the above program instructions, it can execute the above Figure 3 and Figure 7 The method in the corresponding embodiment, therefore, will not be repeated here. For technical details not disclosed in the computer storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application. As an example, the program instructions can be deployed on a computer device, or executed on multiple computer devices located in one location, or, executed on multiple computer devices distributed in multiple locations and interconnected by a communication network.

[0140] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program comprising computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device can perform the above Figure 3 and Figure 7 The method in the corresponding embodiment will therefore not be described in detail here.

[0141] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the above-mentioned program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The above-mentioned storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0142] The above disclosure is only the preferred embodiment of the present application, which certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A signal processing method, characterized in that: The method comprises: Obtaining the audio signal to be processed; Calling the extended enhancement module of the audio signal processing model to perform feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain a target coding feature of the audio signal to be processed, wherein the resolution of the target coding feature is higher than the resolution of the coding feature of the audio signal to be processed; The audio decoding module of the audio signal processing model is called to decode the target coding features of the audio signal to be processed to obtain a target audio signal.

2. The method according to claim 1, characterized in that The calling of the extended enhancement module of the audio signal processing model performs feature enhancement processing on the coding feature of the audio signal to be processed in the time domain to obtain the target coding feature of the audio signal to be processed, further comprising: The audio encoding module of the audio signal processing model is called to perform encoding processing on the audio signal to be processed to obtain the encoding features of the audio signal to be processed.

3. The method according to claim 2, characterized in that The audio encoding module of the audio signal processing model includes M one-dimensional convolution kernels, a regularization function and an activation function, where M is a positive integer; The audio encoding module of the audio signal processing model is called to encode the audio signal to be processed to obtain the encoding feature of the audio signal to be processed, including: Using the M one-dimensional convolution kernels to perform convolution processing on the audio signal to be processed, to obtain a convolution feature vector corresponding to the audio signal to be processed; The convolution feature vector is processed using the regularization function and the activation function to obtain the encoding features of the audio signal to be processed.

4. The method according to any one of claims 1 to 3, characterized in that The coding feature of the audio signal to be processed includes a first unit coding vector and a second unit coding vector; The calling of the extended enhancement module of the audio signal processing model performs feature enhancement processing on the coding features of the audio signal to be processed in the time domain to obtain the target coding features of the audio signal to be processed, including: Inputting the first unit code vector and the initial hidden vector into an extended enhancement module of an audio signal processing model to determine a first hidden vector and a first output vector of the first unit code vector; Inputting the second unit code vector and the first hidden vector into an extended enhancement module of the audio signal processing model to determine a second hidden vector and a second output vector of the second unit code vector; The first output vector and the second output vector are fully connected to obtain target coding features of the audio signal to be processed.

5. The method according to any one of claims 1 to 3, characterized in that The audio decoding module of the audio signal processing model includes N one-dimensional convolution kernels, where N is a positive integer; The calling of the audio decoding module of the audio signal processing model to decode the target coding feature of the audio signal to be processed to obtain the target audio signal includes: The target coding features of the audio signal to be processed are convolved using the N one-dimensional convolution kernels to obtain the target audio signal.

6. The method according to claim 1, characterized in that The method further comprises: Acquire a sample audio set, where the sample audio set includes a plurality of sample audio signals and a reference audio signal corresponding to each sample audio signal, where a resolution of the reference audio signal is greater than a resolution of the sample audio signal; Calling the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal for each sample audio signal; The initial neural network model is trained according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain an audio signal processing model.

7. The method according to claim 6, characterized in that The calling of the initial neural network model to perform signal processing on the sample audio signals in the sample audio set to obtain a predicted audio signal of each sample audio signal includes: Calling the audio encoding module of the initial neural network model to encode the sample audio signal in the sample audio set to obtain a first encoding feature of the sample audio signal; Calling the extended enhancement module of the initial neural network model to perform feature enhancement processing on the first encoding feature of the sample audio signal to obtain a second encoding feature of the sample audio signal; The audio decoding module of the initial neural network model is called to decode the second encoded feature of the sample audio signal to obtain a predicted audio signal of the sample audio signal.

8. The method according to claim 7, characterized in that The step of training the initial neural network model according to the predicted audio signal of each sample audio signal and the reference audio signal of each sample audio signal to obtain an audio signal processing model comprises: adjusting, according to a difference between a predicted audio signal for each sample audio signal and a reference audio signal for each sample audio signal, the audio encoding module of the initial neural network model, and model parameters of the audio encoding module of the initial neural network model; When the adjusted initial neural network model meets the model convergence condition, the initial neural network model after the model parameters are adjusted is used as the audio signal processing model.

9. A computer device, characterized in that: include: storage devices and processors; The storage device stores a computer program; The processor executes the computer program to implement the signal processing method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor and executing the signal processing method according to any one of claims 1 to 8.

Images