Audio processing method and device

Through the special-shaped ring microphone array and neural network processing, the problem that traditional microphone arrays cannot adapt to the special-shaped structure and multi-channel signal processing is difficult to estimate non-stationary noise, achieving more efficient speech processing and real-time rate.

CN112770222BActive Publication Date: 2024-12-03AISPEECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011585187.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-25
Publication Date
2024-12-03
Estimated Expiration
2040-12-25

AI Technical Summary

Technical Problem

Traditional microphone arrays cannot adapt to the structure of special-shaped microphone arrays, and multi-channel signal processing methods are difficult to accurately estimate short-term non-stationary noise.

Method used

Using a special-shaped ring microphone array, 2N audio is formed into two groups of second audio, and N valid audio is obtained through processing, and combined into one third audio for further processing for output. The method includes front-end signal processing and neural network noise reduction, reducing algorithm calculation dimensions, and improving the real-time rate of speech processing.

Benefits of technology

It solves the strict requirements for hardware structure of traditional microphone arrays, makes full use of audio information, improves the real-time rate and signal-to-noise ratio of voice processing, and enhances the voice enhancement effect.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112770222B_ABST
    Figure CN112770222B_ABST
Patent Text Reader

Abstract

The present invention discloses an audio processing method and device, wherein an audio processing method comprises: in response to a shaped ring microphone array acquiring 2N first audio channels, the 2N first audio channels are grouped into two groups of second audio channels, wherein the shaped ring microphone array is a shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1; the N second audio channels are processed separately to acquire valid audio channels in the N second audio channels; the valid audio channels in the two second audio channels are merged into one third audio channel, and the third audio channel is processed and output. By forming a shaped ring microphone array, the strict requirements of the traditional microphone array on the hardware structure can be solved, and further, the acquired audio can be fully utilized, so that the calculation dimension in the algorithm can be reduced and the real-time rate of speech processing can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of audio processing, and in particular to an audio processing method and device. Background Art

[0002] Microphone array solution (commonly used in headsets / mobile phones / conference systems): After obtaining DOA (direction of arrival), the obtained angle information is used to perform beamforming speech enhancement in a specific direction, such as FB (fixed beamforming) and GSC (generalized sidelobecancellation).

[0003] Traditional multi-channel signal processing methods: do not require direct DOA angle information, and use multi-channel signal processing such as BSS (blind source separation), MWF (multichannel wiener filter), MVDR (minimum variance distortionless response), etc.

[0004] The microphone array solution cannot solve the problem of irregular microphones and non-standard microphone array structures;

[0005] Multi-channel signal processing methods cannot accurately estimate short-term non-stationary noise. Summary of the invention

[0006] The embodiments of the present invention provide an audio processing method and device, which are used to solve at least one of the above technical problems.

[0007] In a first aspect, an embodiment of the present invention provides an audio processing method, comprising: in response to an irregular ring microphone array acquiring 2N channels of first audio, the 2N channels of first audio are grouped into two groups of second audio, wherein the irregular ring microphone array is an irregular ring 2N microphone array consisting of one microphone in each of 2N directions, wherein N is a natural number greater than 1; processing the N channels of second audio separately to acquire valid audio in the N channels of second audio; merging the valid audio in the two channels of second audio into one channel of third audio, and processing the third audio and outputting it.

[0008] In a second aspect, an embodiment of the present invention provides an audio processing device, comprising: an acquisition composition program module, configured to, in response to an irregular ring microphone array acquiring 2N channels of first audio, compose the 2N channels of first audio into two groups of second audio, wherein the irregular ring microphone array is an irregular ring 2N microphone array consisting of one microphone in each of 2N directions, wherein N is a natural number greater than 1; a processing acquisition program module, configured to process the N channels of second audio separately to obtain valid audio in the N channels of second audio; a merging processing module, configured to merge the valid audio in the two channels of second audio into one channel of third audio, and output the processed third audio.

[0009] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the audio processing method of any embodiment of the present invention.

[0010] In a fourth aspect, an embodiment of the present invention further provides a computer program product, comprising a computer program stored on a non-volatile computer-readable storage medium, wherein the computer program comprises program instructions, and when the program instructions are executed by a computer, the computer executes the steps of the audio processing method of any embodiment of the present invention.

[0011] The method and device of the present application can solve the strict requirements of the traditional microphone array on the hardware structure by forming a special-shaped ring microphone array. Further, the acquired audio is fully utilized to reduce the calculation dimension in the algorithm and improve the real-time rate of speech processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0013] Figure 1 A flowchart of an audio processing method provided by an embodiment of the present invention;

[0014] Figure 2 A diagram of a special-shaped ring with four microphones as a specific example of an audio processing method provided by an embodiment of the present invention;

[0015] Figure 3 A data flow diagram of a specific example of an audio processing method provided by an embodiment of the present invention;

[0016] Figure 4 An audio processing flow chart of a specific example of an audio processing method provided by an embodiment of the present invention;

[0017] Figure 5 A block diagram of an audio processing device provided by an embodiment of the present invention;

[0018] Figure 6 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Please refer to Figure 1 , which shows a flowchart of an embodiment of an audio processing method of the present application.

[0021] like Figure 1 As shown, in step 101, in response to the special-shaped ring microphone array acquiring 2N first audio channels, the 2N first audio channels are grouped into two groups of second audio channels, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1;

[0022] In step 102, the N channels of second audio are processed respectively to obtain valid audios from the N channels of second audio;

[0023] In step 103, valid audios in the two second audios are combined into one third audio, and the third audio is processed and then output.

[0024] In this embodiment, for step 101, the audio processing device obtains 2N first audio channels in response to the special-shaped ring microphone array, and organizes the 2N first audio channels into two groups of second audio channels, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1. For example, when the special-shaped ring microphone array is placed on the user's waist, there is one microphone in the front, back, left and right to form a special-shaped ring four-microphone array. After obtaining the audio, the audio of two adjacent channels of the four microphone channel audios are organized into two groups of two-channel audios. For example, the four microphones in the front, right, rear and left are MIC1, MIC2, MIC3 and MIC4 respectively. MIC1 and MIC2 can be grouped into one group, and MIC3 and MIC4 can be grouped into one group, or MIC1 and MIC4 can be grouped into one group, and MIC2 and MIC3 can be grouped into one group.

[0025] Afterwards, for step 102, the audio processing device processes the N channels of second audio respectively to obtain valid audio in the N channels of second audio. For example, the audio processing device performs echo cancellation, dereverberation, blind source separation, adaptive beamforming and neural network noise reduction on the N channels of second audio respectively to obtain valid audio in the N channels of second audio.

[0026] Finally, for step 103, the audio processing device combines the valid audio in the two second audio channels into a third audio channel, and processes the third audio channel before outputting it. For example, the third audio channel is again subjected to audio processing such as adaptive beamforming and neural network noise reduction, and finally automatic voice amplification is performed to adjust the final output audio effect.

[0027] The method of this embodiment can solve the strict requirements of the traditional microphone array on the hardware structure by forming a special-shaped ring microphone array. Further, it can make full use of the acquired audio, thereby reducing the calculation dimension in the algorithm and improving the real-time rate of speech processing.

[0028] In the method described in the above embodiment, the step of grouping the 2N first audio channels into two groups of second audio channels includes:

[0029] Split 2N channels of first audio into two groups of N-channel second audio, wherein the two groups of N-channel second audio are both composed of N channels of first audio of adjacent channels. For example, the 2N channels of first audio are four channels of first audio, and the four channels of first audio are MIC1, MIC2, MIC3 and MIC4, wherein MIC1 and MIC2 as well as MIC4 are adjacent, then MIC1 and MIC2 may be grouped together or MIC1 and MIC4 may be grouped together; for example, when MIC1 and MIC2 are grouped together, MIC3 and MIC4 may be grouped together; when MIC1 and MIC4 are grouped together, MIC2 and MIC3 may be grouped together.

[0030] The method of this embodiment splits 2N channels of first audio into two groups of N-channel second audio for processing, thereby making full use of audio information and improving the real-time rate of voice processing.

[0031] In the method described in the above embodiment, the processing of the second audio includes: front-end signal processing and first-level neural network noise reduction.

[0032] In the method described in the above embodiment, the front-end signal processing includes: echo cancellation, dereverberation, blind source separation and first-level automatic beamforming. For example, echo cancellation, dereverberation and blind source separation are used for front-end signal processing to maximize the signal-to-noise ratio, and then the signal-to-noise in the direction of the target human voice is enhanced through automatic beamforming, and then the first-level neural network noise reduction of the above embodiment is used to eliminate the noise component in the audio.

[0033] The method of this embodiment performs front-end signal processing and first-level neural network noise reduction on two channels of audio respectively, thereby making full use of audio information and reducing the calculation dimension in the algorithm.

[0034] In the method described in the above embodiment, the processing of the third audio includes:

[0035] Secondary automatic beamforming, secondary neural network noise reduction and automatic speech gain. For example, the processed second audio is obtained from two sets of audio respectively, and then secondary automatic beamforming and secondary neural network noise reduction are performed. Finally, the audio collected by microphones in multiple directions are synthesized into an audio with a higher signal-to-noise ratio and then automatic speech gain is performed.

[0036] The method of this embodiment performs two-level automatic beamforming, two-level neural network noise reduction and automatic speech gain on the audio, thereby separating the effective audio components in the collected audio to the greatest extent.

[0037] In the method described in the above embodiment, both the primary neural network denoising and the secondary neural network denoising use a language model trained with a clean speech set and a specific noise set to eliminate specific noise.

[0038] In the method described in any of the above embodiments, N is 2, and the special-shaped ring 2N microphone array is a special-shaped ring four-microphone array.

[0039] The method of this embodiment is based on a special-shaped ring four-microphone array, thereby solving the strict requirements of the traditional microphone array on the hardware structure.

[0040] It should be noted that the above method steps are not used to limit the execution order of each step. In fact, some steps may be executed simultaneously or in the opposite order to the specified step order, and this application has no limitation on this.

[0041] The following describes some problems encountered by the inventor in the process of implementing the present invention and a specific embodiment of the solution finally determined, so that those skilled in the art can better understand the solution of the present application.

[0042] The inventors found in the process of implementing this application that the defects in the prior art are mainly caused by the following reasons:

[0043] The microphone array solution uses a standard microphone structure and the inherent phase relationship of multi-channel audio to enhance targets in specific directions and suppress targets in other directions. If the microphone structure cannot meet the algorithm requirements, the algorithm will not take effect.

[0044] Multi-channel signal processing methods, because most of the traditional noise estimation algorithms are based on the assumption that the noise in the analysis time period is more stable than the speech, that is, the noise changes slower than the speech, so it is impossible to accurately and timely estimate the sudden non-stationary noise.

[0045] The inventor also found that there are few special-shaped microphone array products on the market. Most manufacturers have no mature algorithm solutions and sacrifice hardware structure, so they cannot use special-shaped microphone array solutions based on product customization. Algorithm processes and methods are not easy to think of.

[0046] The solution of this application is designed and optimized mainly from the following aspects:

[0047] In order to meet the specific needs of manufacturers, microphone arrays are placed around the waist or in other special application scenarios, with one microphone in each of four directions, forming a special-shaped ring microphone. However, the current standard microphone array obviously cannot meet the needs, and the existing algorithm effect function is very limited. The traditional multi-channel noise reduction algorithm cannot be combined with actual products to achieve better noise reduction effects on audio, and its versatility is poor.

[0048] like Figure 2 , Figure 3 and Figure 4 As shown in the figure, in the application scenario of the special-shaped ring microphone, 2-3 microphones will collect sound source information with a high signal-to-noise ratio. The signal-to-noise ratio between adjacent microphones is relatively close. The algorithm divides the four-channel audio into two groups, and the adjacent channels form two groups of two-channel audio (for example, MIC1 and MIC2, MIC3 and MIC4 are one group respectively). The advantage of this is that it can fully utilize the information of each microphone while reducing the matrix dimension in the subsequent algorithm, reducing the computing power and improving the real-time rate of speech processing.

[0049] The audio composed of each two channels is respectively subjected to echo cancellation, dereverberation, blind source separation, beamforming, etc., to improve the signal-to-noise ratio of the multi-channel noisy conversational speech. The audio obtained from each two channels is subjected to a language model that has been trained with a pure speech set and a specific noise set to eliminate specific noise. The audio output from each group is subjected to front-end signal processing again to improve the signal-to-noise, and then the noise is reduced by the above-mentioned neural network noise reduction model, and finally the processed audio is output through the automatic gain module.

[0050] Through the above description, the effective voice information in the four microphones is synthesized into one, and the secondary signal-to-noise enhancement and noise reduction are performed to ensure the final voice quality.

[0051] Step 1: Channel Grouping:

[0052] Mic1 and mic2, mic3 and mic4 constitute two-channel audio respectively.

[0053] Step 2: Front-end signal processing:

[0054] Including but not limited to echo processing, dereverberation, blind source separation, and automatic beamforming.

[0055] Step 3: Secondary automatic beamforming and neural network noise reduction:

[0056] After the initial noise reduction by the first-level neural network, the second-level automatic beam forming and the second-level neural network noise reduction are carried out.

[0057] Step 4: Automatic Voice Gain:

[0058] Finally, automatic voice gain is performed to adjust the final output audio effect.

[0059] The inventors found that a deeper effect was achieved during the process of implementing the present invention:

[0060] Technical point 1: The present invention is based on a special-shaped ring 4 wheat.

[0061] Effect 1: Solve the strict requirements of traditional microphone arrays on hardware structure.

[0062] Technical point 2: Group the 4 mics.

[0063] Effect 2: Make full use of the voice information in 4mic, reduce the calculation dimension in the algorithm, and improve the real-time rate of voice processing.

[0064] Technical Point 3: Use AEC (acoustic echo cancellation), WPE (dereverberation), and blind source separation (BSS) for front-end signal processing to maximize the signal-to-noise ratio. Then, the signal-to-noise ratio in the target voice direction is enhanced through automatic beamforming. Then, the noise component in the speech is eliminated through a first-level neural network. The audio of one channel is obtained from each of the two sets of results, and then second-order beamforming and neural network noise reduction are performed. Finally, the audio collected by the four mics facing different directions is synthesized into an audio with a high signal-to-noise ratio to the automatic gain module.

[0065] Effect 3: Make full use of the voice information in 4mic, reduce the calculation dimension in the algorithm, and improve the real-time rate of voice processing. The second-level front-end signal processing improves the signal-to-noise ratio and the second-level neural network noise reduction, and separates the effective voice components in the original data collected by 4mic to the greatest extent.

[0066] Please refer to Figure 5 , which shows a block diagram of an audio processing device provided by an embodiment of the present invention.

[0067] like Figure 5 As shown, the audio processing device 500 includes an acquisition composition program module 510, a processing acquisition program module 520 and a merging processing program module 530.

[0068] Among them, the acquisition composition program module 510 is configured to respond to the 2N first audio channels acquired by the special-shaped ring microphone array, and to compose the 2N first audio channels into two groups of second audio channels, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1; the processing acquisition program module 520 is configured to process the N second audio channels separately to obtain valid audio channels in the N second audio channels; the merging processing program module 530 is configured to merge the valid audio channels in the two second audio channels into one third audio channel, and output the processed third audio channel.

[0069] It should be understood that Figure 5 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 5 The modules in it will not be described in detail here.

[0070] It is worth noting that the modules in the embodiments of the present disclosure are not used to limit the scheme of the present disclosure. For example, obtaining the component program module can be described as responding to the 2N-channel first audio obtained by the special-shaped ring microphone array, and composing the 2N-channel first audio into two groups of second audio, wherein the special-shaped ring microphone array is a module of a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1. In addition, the relevant functional modules can also be implemented by a hardware processor, for example, obtaining the component program module block can also be implemented by a processor, which will not be repeated here.

[0071] In some other embodiments, the present invention also provides a non-volatile computer storage medium, the computer storage medium stores computer executable instructions, and the computer executable instructions can execute the audio processing method in any of the above method embodiments;

[0072] As an implementation mode, the non-volatile computer storage medium of the present invention stores computer executable instructions, and the computer executable instructions are configured as follows:

[0073] In response to the special-shaped ring microphone array acquiring 2N first audio channels, the 2N first audio channels are grouped into two groups of second audio channels, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1;

[0074] Process the N channels of second audio respectively to obtain valid audio among the N channels of second audio;

[0075] The valid audios in the two second audios are combined into one third audio, and the third audio is processed and then output.

[0076] The non-volatile computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required by at least one function; the data storage area may store data created according to the use of the audio processing device, etc. In addition, the non-volatile computer-readable storage medium may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the non-volatile computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the audio processing device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0077] An embodiment of the present invention further provides a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer executes any of the above-mentioned audio processing methods.

[0078] Figure 6 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 6 As shown, the device includes: one or more processors 610 and a memory 620, Figure 6 A processor 610 is used as an example. The device of the audio processing method may also include: an input device 630 and an output device 640. The processor 610, the memory 620, the input device 630 and the output device 640 may be connected via a bus or other means. Figure 6 The example of the bus connection is taken. The memory 620 is the above-mentioned non-volatile computer-readable storage medium. The processor 610 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 620, that is, the audio processing method of the above-mentioned method embodiment is realized. The input device 630 can receive input digital or character information, and generate key signal input related to the user setting and function control of the communication compensation device. The output device 640 may include a display device such as a display screen.

[0079] The above product can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0080] As an implementation mode, the electronic device is applied to an audio processing device, used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can:

[0081] In response to the special-shaped ring microphone array acquiring 2N first audio channels, the 2N first audio channels are grouped into two groups of second audio channels, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1;

[0082] Process the N channels of second audio respectively to obtain valid audio among the N channels of second audio;

[0083] The valid audios in the two second audios are combined into one third audio, and the third audio is processed and then output.

[0084] The electronic device of the embodiment of the present application exists in various forms, including but not limited to:

[0085] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0086] (2) Ultra-mobile personal computer devices: These devices fall into the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0087] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0088] (4) Server: A device that provides computing services. The server consists of a processor, hard disk, memory, system bus, etc. The server is similar to a general computer architecture, but because it needs to provide highly reliable services, it has higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0089] (5) Other electronic devices with data interaction functions.

[0090] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0091] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An audio processing method, include: In response to the special-shaped ring microphone array acquiring 2N channels of first audio, the 2N channels of first audio are grouped into two groups of N channels of second audio, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1; the 2N channels of first audio are grouped into two groups of N channels of second audio, comprising: splitting the 2N channels of first audio into two groups of N channels of second audio, wherein the two groups of N channels of second audio are both composed of N channels of first audio of adjacent channels; Processing the two groups of N-channel second audios respectively to obtain valid audios in the two groups of N-channel second audios, wherein processing the two groups of N-channel second audios respectively includes front-end signal processing and primary neural network noise reduction, and the front-end signal processing includes echo cancellation, dereverberation, blind source separation, and primary automatic beamforming; The valid audios in the two groups of N second audios are combined into one third audio, and the third audio is processed and then output.

2. The method according to claim 1, in, The processing of the third audio comprises: Secondary automatic beamforming, secondary neural network noise reduction and automatic speech gain.

3. The method according to claim 2, in, The first-level neural network denoising and the second-level neural network denoising both use a language model trained with a clean speech set and a noise set to eliminate noise.

4. The method according to any one of claims 1 to 3, in, N is 2, and the special-shaped ring 2N microphone array is a special-shaped ring four-microphone array.

5. An audio processing device, include: The acquisition composition program module is configured to, in response to the 2N-channel first audio being acquired by the special-shaped ring microphone array, compose the 2N-channel first audio into two groups of N-channel second audio, wherein the special-shaped ring microphone array is a special-shaped ring 2N microphone array composed of one microphone in each of 2N directions, wherein N is a natural number greater than 1; composing the 2N-channel first audio into two groups of N-channel second audio comprises: splitting the 2N-channel first audio into two groups of N-channel second audio, wherein the two groups of N-channel second audio are both composed of N-channel first audio of adjacent channels; A processing and acquisition program module is configured to process the two groups of N-channel second audios respectively to obtain valid audios in the two groups of N-channel second audios, wherein the processing of the two groups of N-channel second audios respectively includes front-end signal processing and a first-level neural network noise reduction, and the front-end signal processing includes echo cancellation, dereverberation, blind source separation and a first-level automatic beamforming; The merging processing program module is configured to merge the valid audios in the two groups of N second audios into one third audio, and process the third audio and then output it.

6. An electronic device, wherein include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method described in any one of claims 1 to 4.

7. A storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.