WO2021248523A1

WO2021248523A1 - Airflow noise elimination method and apparatus, computer device, and storage medium

Info

Publication number: WO2021248523A1
Application number: PCT/CN2020/096686
Authority: WO
Inventors: 吴锐兴; 田晓晖; 叶利剑
Original assignee: 瑞声声学科技(深圳)有限公司; 瑞声科技(新加坡)有限公司
Priority date: 2020-06-12
Filing date: 2020-06-18
Publication date: 2021-12-16
Also published as: CN111768801A

Abstract

Disclosed in embodiments of the present application is an airflow noise elimination method, comprising: acquiring an original audio signal in a miniature loudspeaker, and preprocessing the original audio signal to obtain an initial signal; performing feature extraction on the initial signal to obtain an audio feature; classifying the initial signal according to the audio feature to determine the type of the initial signal; if the type of the initial signal is a signal containing excitation airflow noise, performing signal compression on the initial signal to obtain a target compressed signal, and analyzing and quantifying the audio feature to determine an audio signal of the airflow noise. Because only the airflow noise signal is filtered out, the original audio signal is retained to the maximum extent; moreover, without changing the structure of the miniature loudspeaker, the sound output quality of the miniature loudspeaker is improved by performing a series of processing on the audio feature, thereby improving the user experience. In addition, also provided are an airflow noise elimination apparatus, a computer device, and a storage medium.

Description

Airflow noise elimination method, device, computer equipment and storage medium

Technical field

This application relates to the field of computer technology, and in particular to a method, device, computer equipment, and storage medium for eliminating airflow noise.

Background technique

Loudspeaker airflow noise is one of the main sources of loudspeaker noise. The cavity of the micro speaker is small and the structure is precise, but the vibration amplitude of the diaphragm is large, and the airflow is easy to form turbulence in the cavity and generate flow-induced noise. The flow-induced noise is amplified by the resonance of a small cavity, and a broadband energy concentration is formed near the high-frequency resonance peak with higher frequency, which forms the airflow noise in people's subjective listening, which is mainly manifested as "hissing" and "sanding" sound. Airflow noise is a common problem in micro-speakers, and it becomes more obvious in the case of large voltage and large amplitude. There are differences between different samples, which are significantly related to the speaker cavity structure and the way of sounding.

technical problem

However, the existing technology often changes the physical structure of the speaker by changing the cavity structure, sound hole and duct design, etc. to improve the airflow noise of the micro speaker. However, this method has a higher process cost, a longer cycle, and is versatile. Sex is limited, and the improvement effect is not good. In view of this, there is an urgent need to provide a new airflow noise elimination method.

Technical solutions

In view of this, the present application provides an airflow noise elimination method, device, computer equipment, and storage medium, which are used to solve the problem of poor airflow noise elimination effect in the prior art.

The specific technical solutions of the embodiments of this application are:

In the first aspect, an embodiment of the present application provides a method for eliminating airflow noise, which is applied to a micro speaker, and includes:

Collecting the original audio signal in the micro speaker, and preprocessing the original audio signal to obtain the initial signal;

Performing feature extraction on the initial signal to obtain audio features;

Classify the initial signal according to the audio feature, and determine the category of the initial signal;

When the type of the initial signal is a signal that includes the noise of the excitation airflow, signal compression processing is performed on the initial signal to obtain a target compressed signal.

Further, after the signal compression processing is performed on the initial signal to obtain the target compressed signal, the method further includes:

The target compressed signal is edited through a fade-in and fade-out mechanism to obtain a target audio signal.

Further, performing feature extraction on the initial signal to obtain audio features includes:

The high-frequency component envelope and the low-frequency component envelope of the initial signal are extracted, and the audio feature is determined according to the ratio of the high-frequency component envelope and the low-frequency component envelope, and/or the initial signal is extracted. The cepstrum coefficient is used as the audio feature.

Further, classifying the initial signal according to the audio feature and determining the category of the initial signal includes:

If the audio feature is determined according to the ratio of the high-frequency component envelope and the low-frequency component envelope, obtaining the ratio of the high-frequency component envelope and the low-frequency component envelope;

When the ratio is less than the preset ratio threshold, it is determined that the type of the initial signal is a signal that includes an exciting airflow noise;

When the ratio is greater than or equal to the preset ratio threshold, it is determined that the type of the initial signal is a signal that does not include exciting airflow noise.

If the audio feature is the Mel cepstrum coefficient of the initial signal, the Mel cepstrum coefficient is input to an audio signal classifier for classification to obtain the category of the initial signal.

Further, performing signal compression processing on the initial signal to obtain a target compressed signal to obtain the target compressed signal includes:

Calculating the diaphragm velocity of the initial signal;

According to the diaphragm speed and a preset speed threshold, signal compression processing is performed on the initial signal to obtain the target compressed signal.

Further, the method further includes:

Acquiring a training sample set, where the training sample set includes the Mel cepstrum coefficient of the initial signal and the corresponding audio category;

Taking the Mel cepstrum coefficient as the input of a preset classifier, and taking the audio category as the desired output, and training the preset classifier to obtain the audio signal classifier that has been trained.

In the second aspect, an embodiment of the present application also provides a device for eliminating airflow noise, including:

The signal acquisition module is used to collect the original audio signal in the micro speaker, and preprocess the original audio signal to obtain the initial signal;

The feature extraction module is used to perform feature extraction on the initial signal to obtain audio features;

A signal classification module, configured to classify the initial signal according to the audio characteristics, and determine the type of the initial signal;

The signal compression module is configured to perform signal compression processing on the initial signal when the type of the initial signal is a signal that includes the noise of the exciting airflow to obtain a target compressed signal.

In the third aspect, the embodiments of the present application also provide a computer device, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, The steps of the method for eliminating air noise as described above are realized.

In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium, including computer instructions, which when run on a computer, cause the computer to execute the steps of the method for eliminating airflow noise as described above.

Beneficial effect

Implementing the embodiments of this application will have the following beneficial effects:

After adopting the above-mentioned airflow noise elimination method, device, computer equipment and storage medium, the original audio signal in the micro speaker is collected, and the original audio signal is preprocessed to obtain the original signal; the original signal is feature extracted to obtain the audio feature; The initial signal is classified according to the audio characteristics, and the category of the initial signal is determined; when the category of the initial signal is a signal that contains exciting airflow noise, signal compression is performed on the initial signal to obtain the target compressed signal, and the audio characteristics are analyzed and quantified , To determine the audio signal of the airflow noise, because only the airflow noise signal is filtered, the original audio signal is retained to the greatest extent, and on the basis of not changing the structure of the micro speaker, a series of audio characteristics are processed to improve Improve the sound output quality of the micro speakers and enhance the user experience.

Description of the drawings

In order to more clearly describe the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only These are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

in:

Figure 1 is a flow chart of the method for eliminating airflow noise in an embodiment;

Fig. 2 is a schematic diagram of a waveform diagram of the target audio signal in an embodiment;

FIG. 3 is a flowchart of a method for determining the type of the initial signal in an embodiment;

Figure 4 is a flowchart of the method for determining the target compressed signal in an embodiment;

Figure 5 is a flow chart of the method for eliminating airflow noise in another embodiment;

FIG. 6 is a schematic diagram of the structure of the method and apparatus for eliminating airflow noise in an embodiment;

Fig. 7 is a schematic diagram of the internal structure of a computer device running the above-mentioned airflow noise elimination method in an embodiment.

Embodiments of the present invention

The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

In order to solve the problem of the poor effect of improving the airflow noise of the micro-speaker by changing the cavity structure, the sound hole and the duct design, etc., the physical structure of the speaker is changed in the traditional technology.

Based on the above problems, in this embodiment, a method for eliminating airflow noise is proposed. The realization of the method can rely on a computer program, which can run on a computer system based on the von Neumann system.

As shown in FIG. 1, the airflow noise elimination method provided in this embodiment is applied to a micro speaker, and the airflow noise elimination method specifically includes the following steps:

Step 102: Collect the original audio signal in the micro speaker, and preprocess the original audio signal to obtain the original signal.

Among them, the original audio signal refers to the unprocessed signal in the WeChat speaker, and the initial signal refers to the audio signal after the original audio signal is preprocessed. Generally, the original audio signal has a small cavity and precise structure due to the micro-speaker, but the amplitude of the diaphragm is large, which makes it easy for the airflow to form turbulence in the cavity and generate flow-induced noise. Therefore, the original audio signal is also pre-processed in this embodiment. Processing to adjust the dynamic range of the original audio signal and the frequency domain energy distribution of the signal, to compensate for the frequency response of the micro-speaker, etc. The preprocessing includes but is not limited to DRC (Dynamic Range Control, dynamic range compression processing), mbDRC (Multi-band Dynamic Range Control, multi-band dynamic range compression processing), EQ (Equaliser, equalization processing) at least one. Understandably, by preprocessing the original audio signal, the original signal is suitable for playback by the micro speaker, so that the subsequent extraction of clearer and more reliable audio features.

Step 104: Perform feature extraction on the initial signal to obtain audio features.

Among them, the audio feature is used to characterize the audio signal. Specifically, the initial signal can be extracted through toolkits such as C/C++, Python, MATLAB, etc., to obtain audio features, such as Mel Cepstral Coefficient (MFCC), linear Prediction general coefficient (LPCC) or component envelope, etc. As the preferred embodiment of this embodiment, in order to better distinguish the airflow noise, the Mel Cepstral Coefficient (MFCC) or component envelope of the initial signal is extracted as the audio feature, so as to make use of the features of visualization and quantization of the audio feature for more accuracy地Analyze.

Step 106: Classify the initial signal according to the audio characteristics, and determine the type of the initial signal.

Specifically, the categories of the initial signal include two categories, namely, the category of signals that include the excitation airflow noise and the category of the signals that do not include the excitation airflow noise. Specifically, by analyzing the audio features, for example, the audio features can be compared with a preset classification threshold, and the category of the initial signal can be determined according to the comparison result; the category classifier of the initial signal can also be pre-trained, and then the audio feature Input to the category classifier, and the output category is the category of the initial signal. Understandably, this embodiment analyzes the audio features and compares the quantized values of the audio features to determine the type of the initial signal, so as to identify the airflow noise contained in the initial signal.

Step 108: When the category of the initial signal is a signal that includes the excitation airflow noise, signal compression processing is performed on the initial signal to obtain a target compressed signal.

Among them, the target compression signal refers to the initial signal that has eliminated the airflow noise. Signal compression is a process used to reduce the vibration speed of audio signals. Specifically, the acoustic model (Speaker Model), such as the Hidden Markov Model (HMM) predicts the intensity of the airflow noise generated by the initial signal on the microspeaker, and reduces the airflow noise of the initial signal to within a preset threshold according to the intensity of the airflow noise. The preset threshold may be different and can be determined by subjective listening. The airflow noise of the initial signal can be slightly compressed by pressing the DRC, and the target compression information can be obtained, and the elimination of the airflow noise can be realized. Understandably, by analyzing and quantifying the audio characteristics, the audio signal of the airflow noise is determined. Since only the airflow noise signal is filtered out, the original audio signal is retained to the greatest extent, and the structure of the micro speaker is not changed. , Through a series of processing of audio characteristics, the sound output quality of the micro speakers is improved, and the user experience is improved.

The above air noise elimination method is to collect the original audio signal in the micro speaker, preprocess the original audio signal to obtain the initial signal; perform feature extraction on the initial signal to obtain the audio feature; classify the initial signal according to the audio feature to determine the initial The type of signal; when the type of the initial signal is a signal that contains exciting airflow noise, the initial signal is compressed to obtain the target compressed signal. By analyzing and quantifying the audio characteristics, the audio signal of the airflow noise is determined. The airflow noise signal is filtered out, the original audio signal is retained to the greatest extent, and on the basis of not changing the structure of the micro speaker, through a series of processing of audio characteristics, the sound output quality of the micro speaker is improved, and the user experience is improved .

In an embodiment, after performing signal compression processing on the initial signal to obtain the target compressed signal, the method further includes:

The target compressed signal is edited through the fade-in and fade-out mechanism to obtain the target audio signal.

Among them, the fade-in and fade-out mechanism is a processing method used to improve the continuity of the audio signal. The fade-in and fade-out mechanism includes fade-in and fade-out. The type of the initial signal changes from a signal that does not contain exciting airflow noise to a signal that contains exciting airflow noise. It works instantaneously, and fade-out works at the moment when the type of the initial signal changes from a signal containing exciting airflow noise to a signal that does not include exciting airflow noise. The target audio signal refers to an audio signal that is played through a micro speaker. Understandably, because step 106 classifies the initial signal, and then performs signal compression processing on the signal that contains the excitation airflow noise, it may result in the signal that contains the excitation airflow noise and the signal that does not contain the excitation airflow. The noise signal has a sudden change, which affects the sound output quality of the micro-speaker. Therefore, the target compressed signal is edited through the fade-in and fade-out mechanism to improve the continuity of the audio signal, thereby improving the quality of the audio signal. As shown in Figure 2, it is the waveform diagram of the target audio signal, where w(t) is the waveform of signal compression processing for the signal containing the excitation airflow noise, f(t) is the waveform of the initial signal category, from It can be seen in Figure 2 that during the time period when the initial signal category changes, the continuity of the target audio signal is achieved by adding a fade-out mechanism, thereby further improving the quality of the target audio signal.

In one embodiment, performing feature extraction on the initial signal to obtain audio features includes:

The high-frequency component envelope and the low-frequency component envelope of the initial signal are extracted, and the audio feature is determined according to the ratio of the high-frequency component envelope and the low-frequency component envelope, and/or the Mel cepstrum coefficient of the initial signal is extracted as the audio feature.

Specifically, the audio feature in this embodiment is the ratio of the high-frequency component envelope and the low-frequency component envelope of the initial signal, and/or the Mel cepstrum coefficient of the initial signal. Among them, the high-frequency component envelope and the low-frequency component envelope can be separately extracted by the square law detection technology (SQL), and then the ratio values of the high-frequency component envelope and the low-frequency component envelope are calculated to obtain the audio characteristics. The Mel cepstrum coefficients can be extracted through the evenlope function or hilbert function that comes with the matlab tool. It is understandable that the ratio of the envelope of the high-frequency component to the envelope of the low-frequency component in this embodiment, and/or the Mel cepstrum coefficient of the initial signal can well reflect the airflow noise information of the audio signal, thus improving The reliability of audio characteristics.

As shown in Fig. 3, in one embodiment, classifying the initial signal according to audio characteristics and determining the category of the initial signal includes:

Step 106A: If the audio feature is determined according to the ratio of the high-frequency component envelope and the low-frequency component envelope, then the ratio of the high-frequency component envelope and the low-frequency component envelope is obtained;

Step 106B: When the ratio is less than the preset ratio threshold, it is determined that the type of the initial signal is a signal that contains exciting airflow noise;

Step 106C: When the ratio is greater than or equal to the preset ratio threshold, it is determined that the type of the initial signal is a signal that does not include the excitation airflow noise.

In this embodiment, when the audio feature is the ratio of the envelope of the high frequency component to the envelope of the low frequency component, by comparing the ratio with a preset ratio threshold, the type of the initial signal is determined according to the comparison result. The preset ratio threshold is a critical value for distinguishing the ratio of the initial signal category. Exemplarily, the preset ratio threshold is 10-5. When the ratio is less than the preset ratio threshold, the type of the initial signal is a signal containing the excitation airflow noise. When the ratio is greater than or equal to the preset ratio threshold, the type of the initial signal It is a signal that does not include exciting airflow noise. It is understandable that by comparing the ratio of the envelope of the high-frequency component and the envelope of the low-frequency component with the preset ratio threshold, the type of the initial signal can be quickly and accurately determined.

In one embodiment, classifying the initial signal according to audio characteristics and determining the category of the initial signal includes:

If the audio feature is the Mel cepstrum coefficient of the initial signal, the Mel cepstrum coefficient is input to the audio signal classifier for classification, and the category of the initial signal is obtained.

In this embodiment, for the case where the audio feature is the Mel cepstrum coefficient of the initial signal, the machine learning model method is used to determine the category of the initial signal, that is, the pre-trained audio signal classifier is used for classification to obtain the initial signal category. The machine learning model can be a support vector machine model SVM, or it can be a classifier set formed by combining multiple weak classifiers using the Ensemble method of ensemble learning. Understandably, the audio signal classifier is used to classify the initial signal. The characteristics of higher accuracy of machine learning are improved, and the accuracy of determining the initial signal category is improved.

As shown in FIG. 4, in one embodiment, performing signal compression processing on the initial signal to obtain the target compressed signal to obtain the target compressed signal includes:

Step 110: Calculate the diaphragm velocity of the initial signal;

Step 112: Perform signal compression processing on the initial signal according to the diaphragm speed and the preset speed threshold to obtain the target compressed signal.

Specifically, the diaphragm speed is an index data used to reflect the intensity of the airflow noise of the audio signal. Specifically, the diaphragm speed is compared with a preset speed threshold. When the diaphragm speed is greater than the preset speed threshold, it is calculated by the acoustic model The diaphragm speed of the initial signal, and then use DRC or mbDRC to compress the initial signal, and compress it to within the preset speed threshold, so as to achieve the function of filtering airflow noise, realize the elimination of airflow noise, and improve The sound quality output by the micro speakers enhances the user experience.

As shown in FIG. 5, in one embodiment, the method for eliminating airflow noise further includes:

Step 114: Obtain a training sample set, the training sample set includes the Mel cepstrum coefficient of the initial signal and the corresponding audio category;

Step 116: Use the Mel cepstrum coefficient as the input of the preset classifier, and use the audio category as the desired output, and train the preset classifier to obtain a trained audio signal classifier.

Specifically, the sample that has been determined to be a signal containing the excitation airflow noise is obtained, and the sample that has been determined to be a signal that does not contain the excitation airflow noise is obtained, and the Mel cepstrum coefficient of the initial signal is used as the preset classifier The audio category is used as the desired output, and the preset classifier is trained to generate the audio category corresponding to the Mel cepstrum coefficient in the training sample set, so as to according to the expected output corresponding to the Mel cepstrum coefficient , Train the preset classifier to get the trained environment classifier.

In this embodiment, the training sample set includes the Mel cepstrum coefficients of the signal containing the excitation airflow noise and the Mel cepstrum coefficients of the signal that does not contain the excitation airflow noise, which ensures the comprehensiveness of the training sample set. In this way, the audio signal classifier trained by the training sample set can learn more comprehensive and accurate audio category classification rules, which improves the efficiency of training the machine learning preset classifier, thereby further improving the efficiency of classifying the initial signal.

Based on the same application concept, an embodiment of the present application provides an airflow noise cancellation device 600, as shown in FIG. 6, including: a signal acquisition module 602, configured to collect the original audio signal in the micro speaker, and preprocess the original audio signal Processing to obtain the initial signal; the feature extraction module 604 is configured to perform feature extraction on the initial signal to obtain audio features; the signal classification module 606 is configured to classify the initial signal according to the audio feature, and determine the initial Signal category; a signal compression module 608, configured to perform signal compression processing on the initial signal to obtain a target compressed signal when the category of the initial signal is a signal containing exciting airflow noise.

Specifically, the airflow noise cancellation device 600 of this embodiment, as shown in FIG. 6, includes: a signal acquisition module 602, configured to collect the original audio signal in the micro speaker, and preprocess the original audio signal to obtain the initial signal The feature extraction module 604 is configured to extract features of the initial signal to obtain audio features; the signal classification module 606 is configured to classify the initial signal according to the audio features and determine the category of the initial signal; The compression module 608 is configured to perform signal compression processing on the initial signal to obtain a target compressed signal when the type of the initial signal is a signal that includes exciting airflow noise. Through the analysis and quantification of the audio characteristics, the audio signal of the airflow noise is determined. Since only the airflow noise signal is filtered, the original audio signal is retained to the greatest extent, and the structure of the micro speaker is not changed. The special feature performs a series of processing to improve the sound output quality of the micro-speaker and enhance the user experience.

It should be noted that the implementation of the airflow noise elimination device in this embodiment is consistent with the realization idea of the above-mentioned airflow noise elimination method, and its implementation principle will not be repeated here. For details, please refer to the corresponding content in the above method.

Fig. 7 shows an internal structure diagram of a computer device in an embodiment. The computer device 700 may specifically be a server or a terminal. As shown in FIG. 7, the computer device includes a processor 710, a memory 720, and a network interface 730 connected through a system bus. Among them, the memory 720 includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can realize the airflow noise elimination method. A computer program can also be stored in the internal memory, and when the computer program is executed by the processor, the processor can execute the airflow noise elimination method. Those skilled in the art can understand that the structure shown in FIG. 7 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may Including more or fewer components than shown in FIG. 7, or combining certain components, or having a different component arrangement.

In an embodiment, the airflow noise elimination method provided by the present application can be implemented in the form of a computer program, and the computer program can be run on a computer device as shown in FIG. 7. The memory of the computer equipment can store various program modules that make up the airflow noise elimination device. For example, the signal acquisition module 602, the feature extraction module 604, the signal classification module 606, and the signal compression module 608.

In one embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the following steps: The original audio signal in the speaker is preprocessed to obtain an initial signal; feature extraction is performed on the initial signal to obtain an audio feature; the initial signal is classified according to the audio feature to determine the The category of the initial signal; when the category of the initial signal is a signal that contains exciting airflow noise, signal compression processing is performed on the initial signal to obtain a target compressed signal.

In one embodiment, a computer-readable storage medium is provided, and the computer-readable storage medium stores a computer program, which is characterized in that, when the computer program is executed by a processor, the following steps are implemented: Original audio signal, preprocess the original audio signal to obtain an initial signal; perform feature extraction on the initial signal to obtain an audio feature; classify the initial signal according to the audio feature to determine the initial signal Category; when the category of the initial signal is a signal that contains exciting airflow noise, signal compression processing is performed on the initial signal to obtain a target compressed signal.

A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer readable storage medium. Here, when the program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and/or volatile memory. Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not a limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The above-disclosed are only the preferred embodiments of the application, and of course the scope of rights of the application cannot be limited by this. Therefore, equivalent changes made in accordance with the claims of the application still fall within the scope of the application.

Claims

A method for eliminating airflow noise, which is characterized in that it is applied to a miniature speaker, and the method includes:

Collecting the original audio signal in the micro speaker, and preprocessing the original audio signal to obtain the initial signal;

Performing feature extraction on the initial signal to obtain audio features;

Classify the initial signal according to the audio feature, and determine the category of the initial signal;

When the type of the initial signal is a signal that includes the noise of the excitation airflow, signal compression processing is performed on the initial signal to obtain a target compressed signal.
The method for eliminating airflow noise according to claim 1, wherein after said performing signal compression processing on said initial signal to obtain a target compressed signal, the method further comprises:

The target compressed signal is edited through a fade-in and fade-out mechanism to obtain a target audio signal.
The method for eliminating airflow noise according to claim 1, wherein said performing feature extraction on said initial signal to obtain audio features comprises:

The high-frequency component envelope and the low-frequency component envelope of the initial signal are extracted, and the audio feature is determined according to the ratio of the high-frequency component envelope and the low-frequency component envelope, and/or the initial signal is extracted. The cepstrum coefficient is used as the audio feature.
The method for eliminating airflow noise according to claim 3, wherein the classifying the initial signal according to the audio characteristics and determining the type of the initial signal comprises:

If the audio feature is determined according to the ratio of the high-frequency component envelope and the low-frequency component envelope, obtaining the ratio of the high-frequency component envelope and the low-frequency component envelope;

When the ratio is less than the preset ratio threshold, it is determined that the type of the initial signal is a signal that includes an exciting airflow noise;

When the ratio is greater than or equal to the preset ratio threshold, it is determined that the type of the initial signal is a signal that does not include exciting airflow noise.
The method for eliminating airflow noise according to claim 3, wherein the classifying the initial signal according to the audio characteristics and determining the type of the initial signal comprises:

If the audio feature is the Mel cepstrum coefficient of the initial signal, the Mel cepstrum coefficient is input to an audio signal classifier for classification to obtain the category of the initial signal.
The method for eliminating airflow noise according to claim 1, wherein the performing signal compression processing on the initial signal to obtain a target compressed signal to obtain a target compressed signal comprises:

Calculating the diaphragm velocity of the initial signal;

According to the diaphragm speed and a preset speed threshold, signal compression processing is performed on the initial signal to obtain the target compressed signal.
The method for eliminating airflow noise according to claim 5, wherein the method further comprises:

Acquiring a training sample set, where the training sample set includes the Mel cepstrum coefficient of the initial signal and the corresponding audio category;

Taking the Mel cepstrum coefficient as the input of a preset classifier, and taking the audio category as the desired output, and training the preset classifier to obtain the audio signal classifier that has been trained.
An airflow noise elimination device, characterized in that the device comprises:

The signal acquisition module is used to collect the original audio signal in the micro speaker, and preprocess the original audio signal to obtain the initial signal;

The feature extraction module is used to perform feature extraction on the initial signal to obtain audio features;

A signal classification module, configured to classify the initial signal according to the audio characteristics, and determine the type of the initial signal;

The signal compression module is configured to perform signal compression processing on the initial signal when the type of the initial signal is a signal that includes the noise of the exciting airflow to obtain a target compressed signal.
A computer device, characterized in that it includes a memory, a processor, and a computer program stored on the memory and capable of running on the processor, wherein the processor implements the rights when the computer program is executed. The steps of the airflow noise elimination method described in any one of 1 to 7 are required.
A computer-readable storage medium, characterized by comprising computer instructions, when the computer instructions run on a computer, causes the computer to execute the steps of the airflow noise elimination method according to any one of claims 1 to 7.