Microphone howling prevention method and device

Abstract

The present invention relates to the field of audio processing, and specifically provides a microphone anti-howling method and device, which is applicable to array microphones, including obtaining a time domain audio signal and generating a corresponding audio spectrum; identifying the howling frequency points on the audio spectrum and responding to the corresponding howling suppression mode; wherein the howling suppression mode includes: an environmental noise howling suppression mode and a self-feedback howling suppression mode; according to the environmental noise howling suppression mode, a feedback-based howling suppression model is constructed, and a first howling suppression strategy is executed; according to the self-feedback howling suppression mode, a directional howling suppression model is constructed, and a second howling suppression strategy is executed. The present application is aimed at the mutual interference problem between the two different howling suppression methods, environmental noise howling suppression and self-feedback howling suppression. Through the automatic response of the howling suppression mode, a dedicated howling suppression model is constructed for the howling caused by environmental noise and the howling of the self-feedback type, and the corresponding howling suppression strategy is executed.

Description

Technical Field

[0001] The present invention relates to the technical field of microphone audio processing, and in particular to a microphone howling prevention method and device. Background Art

[0002] At present, with the rapid development of science and technology, people have higher and higher requirements for the sound quality of microphones. The traditional method of eliminating microphone howling usually detects the frequency corresponding to the maximum value of the real-time received sound and attenuates it. The disadvantage is that the attenuated frequency may be a normal sound signal, and its attenuation will inevitably cause sound distortion. In addition, actual measurements show that the effect of these methods decreases significantly when the gain of the sound amplifier is increased. Therefore, the actual application of the microphone still needs to reasonably arrange the relative position of the speaker and the microphone, and the gain value of the sound amplifier needs to be strictly limited according to the size of the space in the application. If the sound gain value of the amplifier is too large, it is impossible to eliminate the howling by using the traditional howling elimination method in any occasion.

[0003] There are two types of microphone howling: environmentally-induced microphone howling and self-feedback howling. When traditional technical solutions are used to solve different types of howling, when preventing environmentally-induced microphone howling and self-feedback howling, there is a problem that when dealing with one type of howling, the other type of howling will have the opposite effect.

[0004] Here are some anti-squeak measures that may have the opposite effect:

[0005] 1. Volume adjustment:

[0006] For self-feedback howling, lowering the volume can reduce the possibility of feedback.

[0007] However, for environmentally induced howling, if the volume is too low, the desired auditory effect may not be achieved, which may require increasing the volume in some cases, which in turn may aggravate the problem of self-feedback.

[0008] 2. Microphone location:

[0009] To reduce self-feedback, the microphone should be placed away from the speaker to avoid directly capturing the speaker's sound.

[0010] However, if the room acoustics are poor, placing the microphone further away from the speaker may result in capturing more ambient noise, which can exacerbate the howling caused by the environment.

[0011] 3. Room acoustic treatment:

[0012] Using sound-absorbing materials can reduce howling caused by the environment because they can absorb reflected sound and reduce echoes.

[0013] However, excessive sound absorption may make a room feel "dead", and in some cases, appropriate reflections can help distribute the sound evenly in the room and avoid the sound being too concentrated, which can be beneficial when dealing with self-feedback.

[0014] 4. Use feedback suppressor:

[0015] Feedback suppressors can effectively eliminate self-feedback howling, but they may have limited effect on environmentally induced howling, as environmental howling may involve a wider frequency range.

[0016] 5. Use a directional microphone:

[0017] Directional microphones reduce self-feedback because they primarily capture sound in front of you, rather than to the sides or back.

[0018] However, if the ambient noise comes primarily from the sides or rear of the microphone, using a directional microphone may pick up that noise, exacerbating the feedback caused by the environment. Summary of the invention

[0019] The present invention provides a microphone howling prevention method and device, which are used to solve the problem in the above-mentioned background technology.

[0020] In a first aspect, the present invention provides a microphone howling prevention method, which is applicable to an array microphone, comprising:

[0021] Obtain a time-domain audio signal and generate a corresponding audio spectrum;

[0022] Identify the howling frequency points on the audio spectrum and respond to the corresponding howling suppression mode; wherein,

[0023] Howling suppression modes include: environmental noise howling suppression mode and self-feedback howling suppression mode;

[0024] According to the howling suppression mode of the environmental noise, a howling suppression model based on feedback is constructed, and a first howling suppression strategy is executed;

[0025] According to the self-feedback howling suppression mode, a directional howling suppression model is constructed and the second howling suppression strategy is implemented.

[0026] In combination with the first aspect: the acquiring of the time domain audio signal further includes:

[0027] Receive input audio through a microphone, generate an audio array, and determine a frequency domain sequence of each sound source receiving point in the audio array;

[0028] Calculate the sound source delay of each sound source receiving point of the audio array according to the frequency domain sequence and the array structure of the microphone;

[0029] Generate a comparison sequence of input audio according to the time delay of the sound source, and perform fitting optimization of the input audio;

[0030] When the fitting of the input audio is optimized, a time domain conversion is performed to generate a time domain audio signal.

[0031] In combination with the first aspect: the step of identifying the howling frequency points on the audio spectrum and responding to the corresponding howling suppression mode includes:

[0032] Perform cepstrum envelope on the audio spectrum to determine the howling frequency point;

[0033] According to the howling frequency point, the corresponding howling audio feature is determined;

[0034] The howling audio features are processed through a preset howling recognition model to determine the howling cause corresponding to each howling frequency point;

[0035] According to the cause of howling, the preset howling suppression mode is responded.

[0036] In combination with the first aspect: the step of responding to a preset howling suppression mode according to the howling cause includes:

[0037] Calculate the audio frequency difference for each howling frequency point to determine the audio difference value;

[0038] According to the audio difference, the discrete value of the corresponding howling frequency is calculated; where,

[0039] Different howling frequencies correspond to different discrete values ​​in the microphone array;

[0040] The discrete value of each howling frequency point is adapted to the howling discrete intervals in the environmental noise howling suppression mode and the self-feedback howling suppression mode to determine the howling suppression mode.

[0041] In combination with the first aspect: the method of responding to a preset howling suppression mode according to the howling cause further includes:

[0042] Configure the unique suppression index corresponding to the howling frequency and the audio spectrum;

[0043] The suppression index is used to determine the vector relationship between each howling frequency point and the howling suppression mode;

[0044] When there is a correlation between the howling frequency point and any suppression vector of the howling suppression mode, a corresponding suppression point is generated, and the suppression factor is calibrated;

[0045] According to the suppression factor calibration, the corresponding howling suppression mode is determined to call the suppression variable.

[0046] In combination with the first aspect: the construction of the feedback-based howling suppression model includes:

[0047] Obtain sample frequency domain features of ambient noise data and microphone audio feedback;

[0048] Inputting the environmental noise data into the first feedback suppression howling model to obtain the frequency domain characteristics of the environmental noise output by the first feedback suppression howling model;

[0049] Determine the loss function according to the frequency domain characteristics of the environmental noise and the frequency domain characteristics of the sample;

[0050] The suppression calculation layer of the first feedback suppression howling model is updated and trained according to the loss function to generate a second feedback suppression howling model, return to execute the second feedback suppression howling model with input of the ambient noise frequency domain features, and output the first suppression parameter based on the ambient noise howling suppression mode.

[0051] In combination with the first aspect: the construction of the directional howling suppression model includes:

[0052] Obtain microphone directivity data and sample frequency domain features of microphone audio feedback;

[0053] Inputting the microphone directivity data into a third feedback suppression howling model to obtain microphone directivity characteristics output by the third feedback suppression howling model;

[0054] Determine the loss function according to the microphone directivity characteristics and the sample frequency domain characteristics;

[0055] The suppression calculation layer of the third feedback suppression howling model is updated and trained according to the loss function to generate a fourth feedback suppression howling model, return to execute the frequency domain feature of the environmental noise input into the fourth feedback suppression howling model, and output the second suppression parameter based on the self-feedback howling suppression mode.

[0056] In combination with the first aspect: the howling suppression mode corresponding to the response further includes:

[0057] Establishing a feedback mapping node for suppressing feedback corresponding to a howling frequency point and a howling frequency point;

[0058] Acquire the howling frequency suppression mapping information of each frequency point in the audio spectrum according to the feedback mapping node;

[0059] It is determined whether to eliminate howling based on the suppression mapping information.

[0060] In combination with the first aspect: the howling suppression mode corresponding to the response further includes:

[0061] Build an objective function based on the howling suppression mode;

[0062] The objective function is genetically operated with the ideal howling elimination parameter as the maximization suppression target, the target suppression rate of the first howling suppression strategy / the second howling suppression strategy in the current howling suppression mode is determined, and the corresponding suppression strategy is used as the suppression benchmark strategy.

[0063] The beneficial effects of the present invention are:

[0064] First, the present invention can realize howling caused by environmental noise and howling of self-feedback type by automatically responding to the howling suppression mode, build a dedicated howling suppression model, and then execute the corresponding howling suppression strategy. In addition, if the howling caused by environmental noise and the howling of self-feedback type both affect the audio received by the array microphone, two howling suppression modes are responded to at the same time. The present invention is aimed at array microphones, which are a matrix-type sound reception method. The more comprehensive the sound reception, the greater the probability of howling. Moreover, when self-responding, different microphones of the array microphone can realize audio comparison of adjacent microphones, thereby making the howling suppression effect stronger.

[0065] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0066] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0068] Figure 1 A method flow chart of a microphone howling prevention method in an embodiment of the present invention;

[0069] Figure 2 is a flow chart of collecting time domain audio signals in an embodiment of the present invention;

[0070] Figure 3 The figure is a flow chart of howling suppression and elimination detection in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0072] In the prior art, environmental noise howling and self-feedback howling are two different howling phenomena. In the traditional howling suppression or howling attenuation technology, the same method is used to process these two different howling phenomena. However, the howling in the two different situations has different audio spectra. Using the same method to process can only make one type of howling suppression be better; while the other type of howling suppression is processed relatively poorly, and it is impossible to achieve targeted howling suppression, and the suppression effect is naturally poor, but the audio effect is better than the howling of the original sound.

[0073] Therefore, the present invention proposes a microphone howling prevention method, which is applicable to an array microphone, comprising:

[0074] Obtain a time-domain audio signal and generate a corresponding audio spectrum;

[0075] Identify the howling frequency points on the audio spectrum and respond to the corresponding howling suppression mode; wherein,

[0076] Howling suppression modes include: environmental noise howling suppression mode and self-feedback howling suppression mode;

[0077] According to the howling suppression mode of the environmental noise, a howling suppression model based on feedback is constructed, and a first howling suppression strategy is executed;

[0078] According to the self-feedback howling suppression mode, a directional howling suppression model is constructed and the second howling suppression strategy is implemented.

[0079] The principle of the above technical solution is:

[0080] As attached Figure 1 The present application generates a corresponding audio spectrum by acquiring a time-domain audio signal, and the audio spectrum includes but is not limited to audio frequency, amplitude, energy distribution, time, frequency resolution, etc. Due to the time characteristics of the audio signal, it is necessary to sample the time-domain signal, and then convert the signal to the frequency domain through spectrum analysis methods such as Fourier transform, so as to obtain the audio spectrum proposed in the present application, which makes the audio spectrum more accurate.

[0081] Then, the present invention processes the audio spectrum and identifies the howling frequency points therein. The identification methods include but are not limited to scanning and feature recognition. By quickly scanning the audio spectrum, the regions where the frequency, phase and other parameters have changed significantly are found, and it is determined whether these regions are affected by howling, resulting in changes in the audio spectrum.

[0082] Then, this application will identify the howling frequency points according to the changes in the audio spectrum and select the corresponding howling suppression mode. The howling suppression modes mainly include environmental noise howling suppression mode and self-feedback howling suppression mode. The environmental noise howling suppression mode mainly works in a noisy environment and reduces howling through feedback suppression. The self-feedback howling suppression mode works in the audio signal inside the microphone and reduces howling through directionality suppression. After determining the appropriate howling suppression mode, this application will construct a corresponding model based on the characteristics of the mode and execute the corresponding howling suppression strategy.

[0083] The beneficial effects of the above technical solution are:

[0084] First, the present invention can realize howling caused by environmental noise and howling of self-feedback type by automatically responding to the howling suppression mode, build a dedicated howling suppression model, and then execute the corresponding howling suppression strategy. In addition, if the howling caused by environmental noise and the howling of self-feedback type both affect the audio received by the array microphone, two howling suppression modes are responded to at the same time. The present invention is aimed at array microphones, which are a matrix-type sound reception method. The more comprehensive the sound reception, the greater the probability of howling. Moreover, when self-responding, different microphones of the array microphone can realize audio comparison of adjacent microphones, thereby making the howling suppression effect stronger.

[0085] As an embodiment of the present application: the acquiring of the time domain audio signal further includes:

[0086] Receive input audio through a microphone, generate an audio array, and determine a frequency domain sequence of each sound source receiving point in the audio array;

[0087] Calculate the sound source delay of each sound source receiving point of the audio array according to the frequency domain sequence and the array structure of the microphone;

[0088] Generate a comparison sequence of input audio according to the time delay of the sound source, and perform fitting optimization of the input audio;

[0089] When the fitting of the input audio is optimized, a time domain conversion is performed to generate a time domain audio signal.

[0090] The principle of the above technical solution is:

[0091] As attached Figure 2 , this application further optimizes the collected input audio according to the identified howling frequency and the selected howling suppression mode. This step includes two sub-steps:

[0092] The spectrum is adjusted through the audio array; specifically, the audio array returns the frequency domain sequence of each sound source receiving point, and the frequency domain sequence is used to determine the position and strength of the audio received by each microphone (and the howling frequency point), as well as the demand for the howling suppression mode, and the spectrum of the input audio is adjusted. The spectrum of the howling frequency point can be flattened or increased to reduce its impact on the voice signal. At the same time, the spectrum of other non-howling frequency points can be suppressed to prevent the impact of the howling frequency point from spreading to other places.

[0093] Then, through the frequency domain sequence and the overall array structure of the microphone, time domain processing is performed: the audio signal is converted from the frequency domain to the time domain, and the sound source delay is determined. It adapts to the way the human ear perceives sound, because the human ear has different sensitivities to sounds of different frequencies, and in the time domain, these differences can be better processed. At the same time, time domain information can also be used to improve the accuracy of signal processing. In this process, the comparison sequence of the input audio is the audio comparison sequence of different microphones in the microphone array. Then, during the comparison process, the fitting optimization of the input audio is performed to determine the optimal time domain audio signal.

[0094] The beneficial effects of the above technical solution are:

[0095] The present invention can optimize the acquired time-domain audio signal first when performing howling suppression, so that a single audio spectrum can be generated more clearly during the audio signal processing, rather than multiple audio spectrums generated by multiple microphones of the microphone array.

[0096] As an embodiment of the present application: the identifying of the howling frequency points on the audio spectrum and responding to the corresponding howling suppression mode includes:

[0097] Perform cepstrum envelope on the audio spectrum to determine the howling frequency point;

[0098] According to the howling frequency point, the corresponding howling audio feature is determined;

[0099] The howling audio features are processed through a preset howling recognition model to determine the howling cause corresponding to each howling frequency point;

[0100] According to the cause of howling, the preset howling suppression mode is responded.

[0101] The principle of the above technical solution is:

[0102] After determining the howling frequency, the present application processes it through a preset howling recognition model. The preset howling recognition model is a deep learning model, such as a convolutional neural network (CNN) or a recurrent neural network (RNN). The function of the deep learning model is to determine the cause of howling corresponding to each howling frequency. For example, if the howling frequency appears in the low frequency band, it can be determined as the howling caused by environmental noise, and the environmental noise howling suppression mode adopted at this time will take effect. On the contrary, if the howling frequency appears in the high frequency band, it is determined as the howling caused by self-feedback, and the self-feedback howling suppression mode will be adopted at this time. After determining the cause of the howling, the preset howling suppression mode will be responded to.

[0103] When howling suppression is performed, the audio processing algorithm may be changed, or it may be a direct control signal, such as changing the microphone power supply voltage or changing the microphone direction. The goal of this process is to suppress the occurrence of howling by suppressing the cause of howling.

[0104] The beneficial effects of the above technical solution are:

[0105] The present invention can determine the cause of howling by detecting the howling frequency point, thereby responding to the corresponding howling suppression mode.

[0106] As an embodiment of the present application: the step of responding to a preset howling suppression mode according to the howling cause includes:

[0107] Calculate the audio frequency difference for each howling frequency point to determine the audio difference value;

[0108] According to the audio difference, the discrete value of the corresponding howling frequency is calculated; where,

[0109] Different howling frequencies correspond to different discrete values ​​in the microphone array;

[0110] The discrete value of each howling frequency point is adapted to the howling discrete intervals in the environmental noise howling suppression mode and the self-feedback howling suppression mode to determine the howling suppression mode.

[0111] The principle of the above technical solution is:

[0112] When calculating the audio frequency difference of each howling frequency point, the present application obtains the spectrum information of the audio signal through Fourier transform (FFT) or other time domain analysis algorithms, and then calculates the frequency difference for each frequency point to obtain an audio frequency difference matrix. The audio frequency difference matrix contains the frequency difference information of all audio frequency points, which facilitates the calculation of audio discrete values. In the process of calculating each audio element in the audio frequency difference matrix, a mean filter is used to extract the frequency difference value of each frequency point, and the frequency difference value is a discrete value of each howling frequency point. The audio characteristics of different frequency points may be different, so the noise level in different frequency bands determines the corresponding discrete interval. After determining the discrete value of each howling frequency point, it is necessary to adapt the corresponding discrete value interval to the howling discrete interval in the environmental noise howling suppression mode and the self-feedback howling suppression mode. Map each discrete value to a suppression mode that best suits it. This process is based on the nearest neighbor search, genetic algorithm, etc. to calculate the most suitable discrete mapping of the suppression mode. Compare each discrete value with the howling discrete interval under all possible suppression modes to find the most suitable one. Finally, the obtained howling suppression modes are applied to actual audio signal processing to suppress the generated howling.

[0113] The beneficial effects of the above technical solution are:

[0114] When determining the howling suppression mode, the present invention determines the most suitable howling suppression mode for different howling frequency points through discrete calculation.

[0115] As an embodiment of the present application: the step of responding to a preset howling suppression mode according to the howling cause further includes:

[0116] Configure the unique suppression index corresponding to the howling frequency and the audio spectrum;

[0117] The suppression index is used to determine the vector relationship between each howling frequency point and the howling suppression mode;

[0118] When there is a correlation between the howling frequency point and any suppression vector of the howling suppression mode, a corresponding suppression point is generated, and the suppression factor is calibrated;

[0119] According to the suppression factor calibration, the corresponding howling suppression mode is determined to call the suppression variable.

[0120] The principle of the above technical solution is:

[0121] When setting the suppression mode, the variable parameters corresponding to each suppression mode are different. The suppression index is determined by the time domain characteristics of the audio signal and is associated with the frequency response characteristics of the audio spectrum. The suppression index is automatically generated by a neural network or other machine learning algorithm. By determining the vector relationship between the howling frequency and the howling suppression mode, the suppression characteristics of the suppression behavior are determined, thereby calibrating each suppression factor and realizing the call of the suppression variable.

[0122] The beneficial effects of the above technical solution are:

[0123] When adapting the howling frequency point and the howling suppression mode, the present invention determines the audio spectrum of the howling frequency point, sets the howling suppression index, and determines the suppression variable in the corresponding howling suppression mode through the howling suppression index and the corresponding suppression factor, thereby building a corresponding howling suppression model.

[0124] As an embodiment of the present application: the construction of a feedback-based howling suppression model includes:

[0125] Obtain sample frequency domain features of ambient noise data and microphone audio feedback;

[0126] Inputting the environmental noise data into the first feedback suppression howling model to obtain the frequency domain characteristics of the environmental noise output by the first feedback suppression howling model;

[0127] Determine the loss function according to the frequency domain characteristics of the environmental noise and the frequency domain characteristics of the sample;

[0128] The suppression calculation layer of the first feedback suppression howling model is updated and trained according to the loss function to generate a second feedback suppression howling model, return to execute the second feedback suppression howling model with input of the ambient noise frequency domain features, and output the first suppression parameter based on the ambient noise howling suppression mode.

[0129] The principle of the above technical solution is:

[0130] The present application effectively suppresses the howling phenomenon generated by the microphone during the recording process by constructing a feedback suppression howling model. Obtain the sample frequency domain features of the ambient noise data and the microphone audio feedback: collect the ambient noise data and the corresponding microphone audio feedback samples through digital signal processing technology, such as Fourier transform, and convert them into frequency domain features. Input the ambient noise data into the first feedback suppression howling model to obtain the ambient noise frequency domain features output by the first feedback suppression howling model: this is to analyze the ambient noise data using the first feedback suppression howling model and convert the results into frequency domain features. Thereby determining the subsequent loss function and model composition parameters. Determine the loss function based on the ambient noise frequency domain features and the sample frequency domain features: the loss function is used to measure the difference between the microphone audio feedback and the environment. The loss function includes a variety of factors, such as the intensity and frequency distribution of the ambient noise. According to the loss function, the suppression calculation layer of the first feedback suppression howling model is updated and trained to generate a second feedback suppression howling model: this is to use the updated loss function to train the first feedback suppression howling model. The goal of the training is to adjust the parameters of the model so that the model can be as close to the real ambient noise frequency domain features as possible while minimizing the loss. Finally, the frequency domain characteristics of the ambient noise are returned to perform the second feedback suppression.

[0131] The beneficial effects of the above technical solution are:

[0132] The present invention first determines a loss function, and then generates a second feedback suppression howling model under a first feedback suppression howling model through the loss function, so that in the process of building the howling suppression model, the suppression factors are more accurate.

[0133] As an embodiment of the present application: the construction of a directional howling suppression model includes:

[0134] Obtain microphone directivity data and sample frequency domain features of microphone audio feedback;

[0135] Inputting the microphone directivity data into a third feedback suppression howling model to obtain microphone directivity characteristics output by the third feedback suppression howling model;

[0136] Determine the loss function according to the microphone directivity characteristics and the sample frequency domain characteristics;

[0137] The suppression calculation layer of the third feedback suppression howling model is updated and trained according to the loss function to generate a fourth feedback suppression howling model, return to execute the frequency domain feature of the environmental noise input into the fourth feedback suppression howling model, and output the third suppression parameter based on the self-feedback howling suppression mode.

[0138] In the present invention, the directional howling suppression model and the feedback howling suppression model are constructed in a similar manner, so that both can achieve noise suppression more accurately.

[0139] As an embodiment of the present application: the howling suppression mode corresponding to the response further includes:

[0140] Establishing a feedback mapping node for suppressing feedback corresponding to a howling frequency point and a howling frequency point;

[0141] Acquire the howling frequency suppression mapping information of each frequency point in the audio spectrum according to the feedback mapping node;

[0142] It is determined whether to eliminate howling based on the suppression mapping information.

[0143] The principle of the above technical solution is:

[0144] As attached Figure 3 , the present application obtains the suppression mapping information of each howling frequency point in the audio spectrum according to the feedback mapping node, and obtains the corresponding suppression mapping information according to the established feedback mapping relationship between the howling frequency point and the suppression mapping. The suppression mapping information is used to optimize the first feedback suppression howling model, so as to eliminate the howling more effectively. According to the suppression mapping information, it is judged whether the howling is eliminated: this is to judge each howling frequency point in the audio spectrum according to the suppression mapping information provided by the feedback mapping node. If the howling is successfully eliminated, it means that the effect of the first feedback suppression howling model is effective. Otherwise, the model needs to be further adjusted and optimized. By continuously iterating the above steps, the performance of the microphone anti-howling method is gradually improved until the preset standard is met or the optimal solution is reached.

[0145] The beneficial effects of the above technical solution are:

[0146] The present invention can realize continuous iterative optimization of the howling suppression model, thereby achieving the optimal howling elimination effect.

[0147] As an embodiment of the present application: the howling suppression mode corresponding to the response further includes:

[0148] Build an objective function based on the howling suppression mode;

[0149] The objective function is genetically operated with the ideal howling elimination parameter as the maximization suppression target, the target suppression rate of the first howling suppression strategy / the second howling suppression strategy in the current howling suppression mode is determined, and the corresponding suppression strategy is used as the suppression benchmark strategy.

[0150] The principle of the above technical solution is:

[0151] The present application constructs an objective function based on the howling suppression mode: this is to construct a suitable objective function based on the howling suppression mode obtained. The objective function can suppress howling to the maximum extent while minimizing the impact on audio quality. Genetic operations are performed on the objective function with the ideal howling elimination parameter as the maximum suppression target: this is to perform genetic operations on the objective function as a whole. The purpose of this process is to find a set of optimal solutions that can enable the first feedback suppression howling model to achieve the maximum suppression effect without causing too much impact on the audio quality. Determine the target suppression rate of the first howling suppression strategy / second howling suppression strategy under the current howling suppression mode, and use the corresponding suppression strategy as the suppression benchmark strategy: based on the obtained optimal solution, the target suppression rate of the first howling suppression strategy / second howling suppression strategy under the current howling suppression mode can be determined. Then, these optimal suppression strategies can be used as benchmark strategies, and in the subsequent training and optimization process, as comparison and evaluation criteria, to improve the efficiency and accuracy of the entire microphone anti-howling method.

[0152] Based on the same inventive concept, an embodiment of the present invention further provides a microphone anti-howling device, the device comprising:

[0153] Spectrum processing module: obtains the time domain audio signal and generates the corresponding audio spectrum;

[0154] Howling suppression mode response module: identifies the howling frequency points on the audio spectrum and responds to the corresponding howling suppression mode; wherein,

[0155] Howling suppression modes include: environmental noise howling suppression mode and self-feedback howling suppression mode;

[0156] Environmental howling suppression strategy building module: according to the environmental noise howling suppression mode, a feedback-based howling suppression model is built to execute the first howling suppression strategy;

[0157] Self-feedback howling suppression strategy building module: According to the self-feedback howling suppression mode, a directional howling suppression model is built to execute the second howling suppression strategy.

[0158] If the microphone anti-howling device is implemented in software, it can be an access control method program for a mobile storage medium built into the microphone anti-howling device. The microphone anti-howling method program on the microphone anti-howling device completes the operations of mounting a virtual disk, using the virtual disk to access the mobile storage medium, obtaining stored data, processing the stored data, and transmitting files to the target application.

[0159] This application generates the corresponding audio spectrum by acquiring the time domain audio signal. This step is the basis of the entire anti-howling process. Due to the time characteristics of the audio signal, we usually need to sample the time domain signal, and then convert the signal to the frequency domain through spectrum analysis methods such as Fourier transform to obtain the audio spectrum.

[0160] This application will process the audio spectrum and identify the howling frequency points. This is done by quickly scanning the audio spectrum to find areas where parameters such as frequency and phase change significantly. These areas are where howling may occur.

[0161] Then, this application will select the appropriate howling suppression mode according to the identified howling frequency. The howling suppression modes mainly include environmental noise howling suppression mode and self-feedback howling suppression mode. The environmental noise howling suppression mode mainly works in a noisy environment and reduces howling by feedback suppression. The self-feedback howling suppression mode works in the audio signal inside the microphone and reduces howling by directivity suppression.

[0162] After determining a suitable howling suppression mode, the present application will construct a corresponding model according to the characteristics of the mode and execute a corresponding howling suppression strategy.

[0163] Based on the same inventive concept, an embodiment of the present invention provides a microphone anti-howling device, and the microphone anti-howling method program realizes the corresponding functions of the microphone anti-howling method embodiment when executed by the microphone anti-howling device.

[0164] Based on the same inventive concept, an embodiment of the present invention provides a storage medium on which a microphone anti-howling method program is stored. When the microphone anti-howling program is executed by a processor, each step described in any embodiment of the microphone anti-howling method is implemented.

[0165] The storage medium includes volatile or non-volatile, removable or non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, computer program modules or other data). Storage media include, but are not limited to, RAM (Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable read only memory), flash memory or other memory technology, CD-ROM (Compact Disc Read-Only Memory), digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage, or any other medium that can be used to store the desired information and can be accessed by a computer.

[0166] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A microphone howling prevention method, applicable to an array microphone, characterized in that: include: Obtain a time-domain audio signal and generate a corresponding audio spectrum; Identify the howling frequency points on the audio spectrum and respond to the corresponding howling suppression mode; wherein, Howling suppression modes include: environmental noise howling suppression mode and self-feedback howling suppression mode; According to the howling suppression mode of the environmental noise, a howling suppression model based on feedback is constructed, and a first howling suppression strategy is executed; The construction of the feedback-based howling suppression model includes: Obtain sample frequency domain features of ambient noise data and microphone audio feedback; Inputting the environmental noise data into the first feedback suppression howling model to obtain the frequency domain characteristics of the environmental noise output by the first feedback suppression howling model; Determine the loss function according to the frequency domain characteristics of the environmental noise and the frequency domain characteristics of the sample; The suppression calculation layer of the first feedback suppression howling model is updated and trained according to the loss function to generate a second feedback suppression howling model, return to execute the second feedback suppression howling model by inputting the frequency domain features of the environmental noise, and output the first suppression parameter based on the howling suppression mode of the environmental noise; According to the self-feedback howling suppression mode, a directional howling suppression model is constructed to implement the second howling suppression strategy; The construction of the directional howling suppression model includes: Obtain microphone directivity data and sample frequency domain features of microphone audio feedback; Inputting the microphone directivity data into a third feedback suppression howling model to obtain microphone directivity characteristics output by the third feedback suppression howling model; Determine the loss function according to the microphone directivity characteristics and the sample frequency domain characteristics; The suppression calculation layer of the third feedback suppression howling model is updated and trained according to the loss function to generate a fourth feedback suppression howling model, return to execute the frequency domain feature of the environmental noise input into the fourth feedback suppression howling model, and output the third suppression parameter based on the self-feedback howling suppression mode.

2. A microphone howling prevention method as claimed in claim 1, characterized in that: The acquiring of the time domain audio signal further comprises: Receive input audio through a microphone, generate an audio array, and determine a frequency domain sequence of each sound source receiving point in the audio array; Calculate the sound source delay of each sound source receiving point of the audio array according to the frequency domain sequence and the array structure of the microphone; Generate a comparison sequence of input audio according to the time delay of the sound source, and perform fitting optimization of the input audio; When the fitting of the input audio is optimized, a time domain conversion is performed to generate a time domain audio signal.

3. A microphone howling prevention method as claimed in claim 1, characterized in that: The identifying of the howling frequency points on the audio spectrum and responding to the corresponding howling suppression mode includes: Perform cepstrum envelope on the audio spectrum to determine the howling frequency point; According to the howling frequency point, the corresponding howling audio feature is determined; The howling audio features are processed through a preset howling recognition model to determine the howling cause corresponding to each howling frequency point; According to the cause of howling, the preset howling suppression mode is responded.

4. A microphone howling prevention method as claimed in claim 3, characterized in that: The step of responding to a preset howling suppression mode according to the howling cause includes: Calculate the audio frequency difference for each howling frequency point to determine the audio difference value; According to the audio difference, the discrete value of the corresponding howling frequency is calculated; where, Different howling frequencies correspond to different discrete values ​​in the microphone array; The discrete value of each howling frequency point is adapted to the howling discrete intervals in the environmental noise howling suppression mode and the self-feedback howling suppression mode to determine the howling suppression mode.

5. A microphone howling prevention method as claimed in claim 3, characterized in that: The method of responding to a preset howling suppression mode according to the howling cause also includes: Configure the unique suppression index corresponding to the howling frequency and the audio spectrum; The suppression index is used to determine the vector relationship between each howling frequency point and the howling suppression mode; When there is a correlation between the howling frequency point and any suppression vector of the howling suppression mode, a corresponding suppression point is generated, and the suppression factor is calibrated; According to the suppression factor calibration, the corresponding howling suppression mode is determined to call the suppression variable.

6. A microphone howling prevention method as claimed in claim 1, characterized in that: The howling suppression mode corresponding to the response also includes: Establishing a feedback mapping node for suppressing feedback corresponding to a howling frequency point and a howling frequency point; Acquire the howling frequency suppression mapping information of each frequency point in the audio spectrum according to the feedback mapping node; It is determined whether to eliminate howling based on the suppression mapping information.

7. A microphone howling prevention method as claimed in claim 1, characterized in that: The howling suppression mode corresponding to the response also includes: Build an objective function based on the howling suppression mode; The objective function is genetically operated with the ideal howling elimination parameter as the maximization suppression target, the target suppression rate of the first howling suppression strategy / the second howling suppression strategy in the current howling suppression mode is determined, and the corresponding suppression strategy is used as the suppression benchmark strategy.

8. A microphone anti-howling device, applicable to an array microphone, comprising: Spectrum processing module: obtains the time domain audio signal and generates the corresponding audio spectrum; Howling suppression mode response module: identifies the howling frequency points on the audio spectrum and responds to the corresponding howling suppression mode; wherein, Howling suppression modes include: environmental noise howling suppression mode and self-feedback howling suppression mode; Environmental howling suppression strategy building module: according to the environmental noise howling suppression mode, a feedback-based howling suppression model is built to execute the first howling suppression strategy; The construction of the feedback-based howling suppression model includes: Obtain sample frequency domain features of ambient noise data and microphone audio feedback; Inputting the environmental noise data into the first feedback suppression howling model to obtain the frequency domain characteristics of the environmental noise output by the first feedback suppression howling model; Determine the loss function according to the frequency domain characteristics of the environmental noise and the frequency domain characteristics of the sample; The suppression calculation layer of the first feedback suppression howling model is updated and trained according to the loss function to generate a second feedback suppression howling model, return to execute the second feedback suppression howling model by inputting the frequency domain features of the environmental noise, and output the first suppression parameter based on the howling suppression mode of the environmental noise; Self-feedback howling suppression strategy building module: According to the self-feedback howling suppression mode, a directional howling suppression model is built to execute the second howling suppression strategy; The construction of the directional howling suppression model includes: Obtain microphone directivity data and sample frequency domain features of microphone audio feedback; Inputting the microphone directivity data into a third feedback suppression howling model to obtain microphone directivity characteristics output by the third feedback suppression howling model; Determine the loss function according to the microphone directivity characteristics and the sample frequency domain characteristics; The suppression calculation layer of the third feedback suppression howling model is updated and trained according to the loss function to generate a fourth feedback suppression howling model, return to execute the frequency domain feature of the environmental noise input into the fourth feedback suppression howling model, and output the third suppression parameter based on the self-feedback howling suppression mode.

Images