An audio signal compensation method based on AI control and related device

Abstract

The present invention discloses an audio signal compensation method and related devices based on AI control, which relates to the field of artificial intelligence technology. The method includes: extracting features of an audio signal to be compensated to obtain audio feature information; calculating the spectrum fluctuation of the audio signal to be compensated, identifying the audio category based on the spectrum fluctuation and the audio feature information, and obtaining the corresponding audio category; obtaining user habit information corresponding to the audio category, and determining a first compensation parameter based on the user habit information by using equalizer compensation analysis; obtaining frequency response data, and generating a target correction coefficient based on the frequency response data combined with the environmental frequency response parameter by using impulse response analysis to obtain a second compensation parameter; compensating the audio signal to be compensated based on the second compensation parameter, and transmitting the target audio signal to a speaker for playback. The present invention improves the user's auditory experience and enables audio signal compensation to achieve a more ideal effect.

Description

Technical Field

[0001] The present invention mainly relates to the field of audio processing technology, and in particular to an audio signal compensation method based on AI control and related devices. Background Art

[0002] With the rapid development of acoustic technology, people are paying more and more attention to the audio effects of sound systems. In order to improve the quality of audio playback, audio signals are usually compensated to meet people's desired auditory effects. Audio category recognition of audio signals is an important part of audio signal compensation methods. At present, audio category recognition of audio signals is usually divided into audio categories by labeling by relevant personnel, but this method is too dependent on the professional qualities of relevant personnel and cannot guarantee the accuracy of audio category recognition. At the same time, the human cost invested is too high. In the current audio signal compensation, user habits are rarely considered in the generation of audio signal compensation parameters, resulting in the final compensated audio still needing to be adjusted by the user, resulting in insufficient intelligence of audio signal compensation, and unable to bring users a more convenient audio playback experience. At the same time, in the current audio signal compensation, since the frequency response of the speaker itself is not flat enough and there are problems such as audio signal reflection and absorption in the audio playback space, in order to solve these problems, the audio playback effect is usually improved by adding sound-absorbing materials in the audio playback space. However, this method is costly and has poor compensation effect on the audio signal, resulting in the audio signal compensation failing to achieve the expected effect and failing to provide users with a high-quality audio playback experience. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art. The present invention provides an audio signal compensation method and related devices based on AI control, which improves the user's auditory experience and enables the audio signal compensation to achieve a more ideal effect.

[0004] In order to solve the above technical problems, the present invention provides an audio signal compensation method based on AI control, which is applied to an audio pickup, an embedded real-time operating system and a speaker, wherein the embedded real-time operating system is connected to the audio pickup and the speaker respectively; the method comprises:

[0005] Acquire an audio signal to be compensated based on an audio pickup, and transmit the audio signal to be compensated to an embedded real-time operating system, wherein the embedded real-time operating system extracts features of the audio signal to be compensated through an artificial intelligence engine to obtain audio feature information;

[0006] Calculating the frequency spectrum fluctuation of the audio signal to be compensated, and performing audio category recognition on the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information to obtain an audio category corresponding to the audio signal to be compensated;

[0007] Acquire user habit information corresponding to the audio category, and determine a first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information;

[0008] Acquire frequency response data of the audio signal to be compensated, generate a target correction coefficient by using impulse response analysis based on the frequency response data combined with an environmental frequency response parameter, and adjust the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter;

[0009] The audio signal to be compensated is compensated based on the second compensation parameter to obtain a target audio signal, and the target audio signal is transmitted to a speaker, and the speaker plays the target audio signal.

[0010] Optionally, the extracting features of the audio signal to be compensated to obtain audio feature information includes:

[0011] Preprocessing the audio signal to be compensated to obtain a preprocessed audio signal to be compensated;

[0012] Performing segmentation processing on the preprocessed audio signal to be compensated to obtain the segmented audio signal to be compensated;

[0013] Extracting short-time audio features from the audio signal to be compensated after segmentation to obtain short-time audio feature information;

[0014] The long-term audio feature is extracted from the audio signal to be compensated after the segmentation process to obtain long-term audio feature information, and audio feature information is generated based on the short-term audio feature information and the long-term audio feature information.

[0015] Optionally, calculating the spectrum fluctuation of the audio signal to be compensated, and identifying the audio category of the audio signal to be compensated based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated includes:

[0016] Obtaining the kurtosis of the high frequency band of the frequency spectrum of each audio frame in the audio signal to be compensated;

[0017] Acquire the frequency spectrum of each audio frame in the audio signal to be compensated, and calculate the logarithmic energy difference of the frequency spectrum of each audio frame in the mid- and low-frequency band frequency spectrum;

[0018] Acquire Mel-frequency cepstral coefficients of each audio frame in the audio signal to be compensated, perform natural logarithm calculation processing based on the Mel-frequency cepstral coefficients to obtain a logarithmic energy vector, and generate energy fluctuation data based on the logarithmic energy vector;

[0019] generating spectrum fluctuations based on the spectrum high frequency band kurtosis, logarithmic energy difference and energy fluctuation data;

[0020] The spectrum fluctuation and audio feature information are input into an audio classification model. The audio classification model performs audio category recognition based on the spectrum fluctuation and audio feature information to obtain an audio category corresponding to the audio signal to be compensated, wherein the audio classification model includes a residual gated convolutional neural network based on a channel attention mechanism and a bidirectional long short-term memory network.

[0021] Optionally, determining the first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information includes:

[0022] Determining a corresponding compensation filter based on the audio category, and acquiring a first gain coefficient of the compensation filter based on the user habit information;

[0023] Calculate a timbre gain factor and a volume gain factor based on the user habit information and the audio feature information;

[0024] Acquire a frequency distribution curve of the audio signal to be compensated, determine an amplitude value of each frequency point in the frequency distribution curve, and determine a second gain coefficient based on the user habit information and the amplitude value of each frequency point;

[0025] An equalizer gain coefficient is calculated based on the first gain coefficient and the second gain coefficient, and a first compensation parameter of the audio signal to be compensated is determined based on the equalizer gain coefficient, a timbre gain factor and a volume gain factor.

[0026] Optionally, the acquiring frequency response data of the audio signal to be compensated includes:

[0027] Performing short-time Fourier transform on the audio signal to be compensated to obtain short-time frequency domain information;

[0028] Calculating an autocorrelation function of the audio signal to be compensated based on the short-time frequency domain information, and determining a linear prediction error filter coefficient based on the autocorrelation function;

[0029] Frequency response calculation is performed based on the linear prediction error filter coefficient using a sliding window and a random suppression curve to obtain frequency response data.

[0030] Optionally, generating a target correction coefficient by using impulse response analysis based on the frequency response data combined with environmental frequency response parameters includes:

[0031] The embedded real-time operating system controls the audio pickup to collect the test audio signal of the current environment, and obtains the audio parameters of the test audio signal;

[0032] Generate environmental frequency response parameters based on audio parameters of the test audio signal combined with audio parameters of a preset standard audio signal;

[0033] Acquire the peak occurrence time of the test audio signal, perform impulse response analysis based on the peak occurrence time in combination with a preset head-related transfer function, and obtain impulse response data;

[0034] A target correction coefficient is obtained by performing calculation based on the frequency response data, the environmental frequency response parameters and the impulse response data.

[0035] Optionally, compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal includes:

[0036] The audio signal to be compensated is dynamically corrected based on the second compensation parameter in combination with the filter coefficient to obtain a target audio signal.

[0037] In addition, the present invention also provides an audio signal compensation device based on AI control, which is applied to an audio pickup, an embedded real-time operating system and a speaker, wherein the embedded real-time operating system is connected to the audio pickup and the speaker respectively; the device comprises:

[0038] Audio feature extraction module: used to obtain the audio signal to be compensated based on the audio pickup, and transmit the audio signal to be compensated to the embedded real-time operating system, and the embedded real-time operating system extracts the features of the audio signal to be compensated through the artificial intelligence engine to obtain audio feature information;

[0039] An audio category recognition module is used to calculate the frequency spectrum fluctuation of the audio signal to be compensated, and to identify the audio category of the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information, so as to obtain the audio category corresponding to the audio signal to be compensated;

[0040] A first compensation parameter acquisition module: used to acquire user habit information corresponding to the audio category, and determine the first compensation parameter of the audio signal to be compensated by using equalizer compensation analysis based on the user habit information;

[0041] A second compensation parameter acquisition module is used to acquire frequency response data of the audio signal to be compensated, generate a target correction coefficient based on the frequency response data combined with the ambient frequency response parameters by using impulse response analysis, and adjust the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter;

[0042] Audio signal compensation module: used for compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal, and transmitting the target audio signal to a speaker, and the speaker plays the target audio signal.

[0043] In addition, the present invention also provides an audio signal compensation system based on AI control, the system comprising an audio pickup, an embedded real-time operating system and a speaker, the embedded real-time operating system being connected to the audio pickup and the speaker respectively, and the system being configured to execute the above-mentioned audio signal compensation method based on AI control.

[0044] In addition, the present invention also provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the above-mentioned audio signal compensation method based on AI control.

[0045] In the embodiment of the present invention, the audio category of the audio signal to be compensated is identified based on the spectrum fluctuation and audio feature information, which can improve the accuracy of audio category identification and greatly reduce the investment of manpower cost. The first compensation parameter of the audio signal to be compensated is determined by equalizer compensation analysis based on user habit information. Taking into account the influence of user habit information on audio signal compensation, the intelligence of audio signal compensation is improved, so that the final played audio is more in line with the user's daily habits. The frequency response data of the audio signal to be compensated is obtained, and the target correction coefficient is generated by impulse response analysis based on the frequency response data combined with the environmental frequency response parameters to adjust the first compensation parameter and generate the second compensation parameter. By analyzing the frequency response data of the audio signal to be compensated, the audio output can be made flatter. At the same time, the final compensation parameter is generated by impulse response analysis combined with the environmental frequency response parameters, which can avoid the interference of the spatial environment on the audio signal to the greatest extent. The audio signal to be compensated is compensated by the second compensation parameter to generate the target audio signal, so as to better realize the intelligent control of the audio signal, so that the final played audio can be closer to the ideal playback effect, improve the user's auditory experience, and achieve a more ideal effect of audio signal compensation. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1is a flow chart of an audio signal compensation method based on AI control in an embodiment of the present invention;

[0048] Figure 2 is a flow chart of an audio signal compensation method based on AI control in another embodiment of the present invention;

[0049] Figure 3 Schematic diagram of the structure of an audio signal compensation system based on AI control in an embodiment of the present invention;

[0050] Figure 4 It is a schematic diagram of the structural composition of an audio signal compensation device based on AI control in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0052] Embodiment 1

[0053] See also Figure 1 , Figure 1 : is a flow chart of an audio signal compensation method based on AI control in an embodiment of the present invention, the method is applied to an audio pickup, an embedded real-time operating system and a speaker, the embedded real-time operating system is connected to the audio pickup and the speaker respectively; the method comprises:

[0054] S11: acquiring an audio signal to be compensated based on an audio pickup, and transmitting the audio signal to be compensated to an embedded real-time operating system, wherein the embedded real-time operating system extracts features of the audio signal to be compensated through an artificial intelligence engine to obtain audio feature information;

[0055] In the specific implementation process of the present invention, the feature extraction of the audio signal to be compensated to obtain audio feature information includes: preprocessing the audio signal to be compensated to obtain the preprocessed audio signal to be compensated; segmenting the preprocessed audio signal to be compensated to obtain the segmented audio signal to be compensated; extracting short-time audio features of the segmented audio signal to be compensated to obtain short-time audio feature information; extracting long-time audio features of the segmented audio signal to be compensated to obtain long-time audio feature information, and generating audio feature information based on the short-time audio feature information and the long-time audio feature information.

[0056] Specifically, the audio signal to be compensated is obtained through an audio pickup, and the audio signal to be compensated is transmitted to the embedded real-time operating system through the communication connection between the audio pickup and the embedded real-time operating system. The embedded real-time operating system has a built-in artificial intelligence (AI) engine, and the AI ​​engine can perform relevant operations and processes for audio signal compensation to realize AI control of audio signal compensation. First, the audio signal to be compensated is preprocessed, and the preprocessing includes pre-emphasis processing, windowing processing, and analog-to-digital conversion processing, etc., to obtain the preprocessed audio signal to be compensated. The preprocessed audio signal to be compensated is segmented according to a preset interval to obtain the segmented audio signal to be compensated. Short-time audio feature extraction is performed on the segmented audio signal to be compensated, and short-time audio feature extraction is performed through a feature extraction model to obtain short-time audio feature information, and the short-time audio feature information includes time domain features, frequency domain features, and cepstrum features. Long-time audio feature extraction is performed on the audio signal to be compensated after segmentation processing, and Gaussian supervector feature extraction is performed on the audio signal to be compensated after segmentation processing through a Gaussian mixture model to obtain Gaussian supervector feature information, and the Gaussian supervector feature information is mapped to a low-dimensional space for compression projection to obtain long-time audio feature information. The short-time audio feature information and the long-time audio feature information constitute audio feature information. The combination of the long-time audio feature information and the short-time audio feature information can represent complex audio information, thereby obtaining more comprehensive audio feature information.

[0057] S12: calculating the frequency spectrum fluctuation of the audio signal to be compensated, and performing audio category recognition on the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information to obtain an audio category corresponding to the audio signal to be compensated;

[0058] In a specific implementation process of the present invention, the spectrum fluctuation of the audio signal to be compensated is calculated, and the audio category of the audio signal to be compensated is identified based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated, including: obtaining the high-frequency band kurtosis of the spectrum of each audio frame in the audio signal to be compensated; obtaining the spectrum of each audio frame in the audio signal to be compensated, and calculating the logarithmic energy difference of the spectrum of each audio frame in the mid- and low-frequency band spectrum; obtaining the Mel-frequency cepstral coefficients of each audio frame in the audio signal to be compensated, performing natural logarithm calculation processing based on the Mel-frequency cepstral coefficients to obtain a logarithmic energy vector, and generating energy fluctuation data based on the logarithmic energy vector; generating spectrum fluctuation based on the high-frequency band kurtosis of the spectrum, the logarithmic energy difference and the energy fluctuation data; inputting the spectrum fluctuation and audio feature information into an audio classification model, and the audio classification model performs audio category identification based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated, wherein the audio classification model includes a residual gated convolutional neural network based on a channel attention mechanism and a bidirectional long short-term memory network.

[0059] Specifically, the high-band kurtosis of the spectrum of each audio frame in the audio signal to be compensated is obtained. The high-band kurtosis of the spectrum represents the kurtosis or energy sharpness of the spectrum of the audio frame in the high-frequency band. By calculating the local peak of the frequency point of the audio frame, the local valley of the spectrum closest to the high-frequency side and the low-frequency side of the frequency point is calculated. The high-band kurtosis of the spectrum can be obtained according to the local peak and the local valley of the spectrum. The spectrum of each audio frame in the audio signal to be compensated is obtained, and the logarithmic energy difference of the spectrum of each audio frame in the mid-low frequency band spectrum is calculated, that is, the logarithmic energy difference of the corresponding frequency between adjacent audio frames in the mid-low frequency band spectrum is calculated. The Mel-frequency cepstral coefficient of each audio frame in the audio signal to be compensated is obtained, and each audio frame in the audio signal to be compensated is fast Fourier transformed to obtain the corresponding spectrum function. The spectrum function is passed through a Mel filter group composed of several Mel filters, and the energy value of the spectrum frame within the spectrum range of each filter is calculated. All energy values ​​constitute the Mel-frequency cepstral coefficient. The Mel-frequency cepstral coefficient can reflect the distribution of the energy of the audio signal in different frequency ranges. Based on the Mel frequency cepstrum coefficient, natural logarithm calculation is performed, and the low eight-dimensional energy value in the Mel frequency cepstrum coefficient is taken to perform natural logarithm calculation to obtain the logarithmic energy vector, and energy fluctuation data is generated based on the logarithmic energy vector, and the variance value and mean value are calculated based on the logarithmic energy vector of each audio frame, and the Mel energy fluctuation feature is formed according to the variance value and mean value, that is, the energy fluctuation data. Based on the kurtosis of the high frequency band of the spectrum, the logarithmic energy difference and the energy fluctuation data, the spectrum fluctuation is generated, that is, the spectrum fluctuation is calculated according to the value of the high frequency band of the spectrum, the logarithmic energy difference and the energy fluctuation data and the corresponding weight coefficient. The spectrum fluctuation is used to represent the energy fluctuation of the spectrum of the audio signal, which can reflect the activity of the sound in the audio. If the spectrum fluctuation is high, the audio signal may be percussion music, which is information that cannot be represented by general audio features. The kurtosis of the high frequency band of the spectrum can reflect the energy in the high frequency band, the logarithmic energy difference can reflect the energy in the middle and low frequency bands, and the energy fluctuation data reflects the Mel energy distribution range, so the spectrum fluctuation obtained is more specific and accurate.The spectrum fluctuation and audio feature information are input into an audio classification model, and the audio classification model performs audio category recognition based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated, wherein the audio classification model includes a residual gated convolutional neural network and a bidirectional long short-term memory network based on a channel attention mechanism. The residual gated convolutional neural network based on the channel attention mechanism includes two residual gated convolutional units, a Sequeeze-and-Excitation (SE) structure and a maximum pooling layer. The residual gated convolutional neural network is used to extract deep local features, and the bidirectional long short-term memory network is used to summarize time domain information. At the same time, the channel attention mechanism can be combined to assign different attention weights to the output of the neural network, thereby improving the performance of the model for audio classification, thereby improving the accuracy of audio category recognition and greatly reducing the investment in manpower costs.

[0060] S13: Acquire user habit information corresponding to the audio category, and determine a first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information;

[0061] In the specific implementation process of the present invention, the first compensation parameter of the audio signal to be compensated is determined by using the equalizer compensation analysis based on the user habit information, including: determining the corresponding compensation filter based on the audio category, and obtaining the first gain coefficient of the compensation filter based on the user habit information; calculating the timbre gain factor and the volume gain factor based on the user habit information combined with the audio feature information; obtaining the frequency distribution curve of the audio signal to be compensated, determining the amplitude value of each frequency point in the frequency distribution curve, and determining the second gain coefficient based on the user habit information combined with the amplitude value of each frequency point; calculating the equalizer gain coefficient based on the first gain coefficient and the second gain coefficient, and determining the first compensation parameter of the audio signal to be compensated based on the equalizer gain coefficient, the timbre gain factor and the volume gain factor.

[0062] Specifically, the user habit information corresponding to the audio category is obtained from the database. The user habit information includes habitual volume and habitual sound quality, etc. Users of different audio categories have different playing habits. The corresponding compensation filter is determined based on the audio category. Different audio categories correspond to different compensation filters. The compensation filter may include a bandpass filter, a loan filter, and a finite length unit impulse response filter, etc., and the first gain coefficient of the compensation filter is obtained based on the user habit information, that is, the gain coefficient of the compensation filter is calculated according to the user habit information. The gain coefficient is the gain weight corresponding to the frequency component to be compensated in the audio signal. The timbre gain factor and the volume gain factor are calculated based on the user habit information combined with the audio feature information, the user habit volume level is extracted from the user habit information, and the volume gain factor is calculated according to the user habit volume level. The user habit sound quality is extracted from the user habit information, and the tone gain and the harmonic component gain are calculated according to the user habit sound quality combined with the audio feature information. The timbre gain factor is generated by the tone gain and the harmonic component gain. Obtain the frequency distribution curve of the audio signal to be compensated, obtain the frequency of the sine wave component of the audio signal to be compensated, determine the amplitude and phase of the sine wave component frequency, expand it in frequency order according to the amplitude and phase, obtain the frequency function, analyze it in the frequency domain according to the frequency function, and obtain the frequency distribution curve. Determine the amplitude value of each frequency point in the frequency distribution curve, and determine the second gain coefficient based on the user habit information and the amplitude value of each frequency point, determine the adjustment coefficient of the amplitude value of each frequency point according to the user habit information, and determine the second gain coefficient according to the adjustment coefficient. Calculate the equalizer gain coefficient based on the first gain coefficient and the second gain coefficient, calculate the first gain coefficient and the second gain coefficient and their corresponding weights to obtain the equalizer gain coefficient, and determine the first compensation parameter of the audio signal to be compensated based on the equalizer gain coefficient, the timbre gain factor and the volume gain factor, that is, determine the compensation parameters of the equalizer and the filter for the audio signal to be compensated.

[0063] S14: acquiring frequency response data of the audio signal to be compensated, generating a target correction coefficient by using impulse response analysis based on the frequency response data combined with an environmental frequency response parameter, and adjusting the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter;

[0064] In the specific implementation process of the present invention, the obtaining of the frequency response data of the audio signal to be compensated includes: performing a short-time Fourier transform on the audio signal to be compensated to obtain short-time frequency domain information; calculating the autocorrelation function of the audio signal to be compensated based on the short-time frequency domain information, and determining a linear prediction error filter coefficient based on the autocorrelation function; and performing a frequency response calculation based on the linear prediction error filter coefficient using a sliding window and a random suppression curve to obtain frequency response data.

[0065] Furthermore, the method of generating a target correction coefficient by using impulse response analysis based on the frequency response data in combination with environmental frequency response parameters includes: the embedded real-time operating system controls the audio pickup to collect a test audio signal of the current environment and obtains audio parameters of the test audio signal; generates environmental frequency response parameters based on the audio parameters of the test audio signal in combination with audio parameters of a preset standard audio signal; obtains a peak occurrence time of the test audio signal, performs impulse response analysis based on the peak occurrence time in combination with a preset head-related transfer function, and obtains impulse response data; and performs calculations based on the frequency response data, environmental frequency response parameters, and impulse response data to obtain a target correction coefficient.

[0066] Specifically, the audio signal to be compensated is subjected to a short-time Fourier transform to obtain short-time frequency domain information, that is, to obtain information in the short-time frequency domain. The autocorrelation function of the audio signal to be compensated is calculated based on the short-time frequency domain information, and the square sum of the real part and the imaginary part of each frequency point in the audio signal to be compensated is calculated according to the short-time frequency domain information to determine the target envelope of the audio signal to be compensated after segmentation processing. The envelope of the audio refers to the control of various audio parameters through a function curve to achieve dynamic changes in volume and other effects. According to the target envelope, the envelope of the audio signal to be compensated after segmentation processing is flattened to obtain the audio signal to be compensated after the envelope flattening. According to the square sum, the audio signal to be compensated after the envelope flattening is subjected to an inverse fast Fourier transform to obtain the autocorrelation function, and the linear prediction error filter coefficient is determined based on the autocorrelation function. The autocorrelation function is solved and calculated using the Levinson-Dubin algorithm, which is an algorithm for linear prediction and calculation of autocorrelation values, to obtain the linear prediction coefficient, and the linear prediction error filter coefficient is generated according to the linear prediction coefficient. Based on the linear prediction error filter coefficient, the frequency response is calculated using a sliding window and a random suppression curve. By simulating the audio time domain signal after the frequency response, the enhancement strength of the audio signal is determined. The random suppression curve is determined according to the enhancement strength. The amplitude jump is smoothed by the sliding window, so that the amplitude difference between each frequency point of the audio signal is readjusted. The frequency response is calculated by combining the linear prediction error filter coefficient with the amplitude jump smoothing data obtained by the sliding window and the enhancement strength obtained by the random suppression curve to obtain frequency response data. The frequency response data captures the relative difference in amplitude of the sound signal at different frequencies. The true coefficient generated by the frequency response data can better adjust the flatness of the audio signal. The embedded real-time operating system controls each speaker to play the test audio in the current environment in turn. The embedded real-time operating system controls the audio pickup to collect the test audio signal of the current environment. The test audio signal is generated by the current environment, and the audio parameters of the test audio signal, such as loudness, are obtained. The environmental frequency response parameters are generated based on the audio parameters of the test audio signal combined with the audio parameters of the preset standard audio signal. The preset standard audio signal is an audio signal without environmental absorption. The audio parameters of the test audio signal are compared with the audio parameters of the preset standard audio signal. The difference frequency band between the two frequency response parameters is determined according to the comparison result. The environmental frequency response parameters are generated according to the difference frequency band. The environmental frequency response parameters can be used to correct the impact of the current environment on the audio signal playback.The peak occurrence time of the test audio signal is obtained, and impulse response analysis is performed based on the peak occurrence time combined with a preset head-related transfer function. A preset user model is extracted from the simulation software, and the user model includes a head model and a trunk model. The user model is used in combination with the Green function to perform multiple scattering calculations through the finite difference method to obtain the head-related transfer function, and the first impulse response data of the high-frequency band of the head-related transfer function is obtained. The peak occurrence time of each test audio signal is used for calculation through the truncation method to obtain the second impulse response data in the current environment, and the final impulse response data is generated from the first impulse response data and the second impulse response data. The reflected sound data and the head-related impulse response of the audio in the playback space can be obtained through impulse response analysis, and the sound coloration problem of the current environment on the audio playback can be obtained. The frequency response data, the environmental frequency response parameters and the impulse response data are calculated based on the frequency response data, the environmental frequency response parameters and the impulse response data, and the frequency response data, the environmental frequency response parameters and the impulse response data are input into the correction coefficient calculation model for calculation to obtain the target correction coefficient. The first compensation parameter is adjusted based on the target correction coefficient, and the filter coefficient and the equalization coefficient in the first compensation parameter are adjusted according to the target correction coefficient to obtain the second compensation parameter.

[0067] S15: Compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal, and transmitting the target audio signal to a speaker, so that the speaker plays the target audio signal.

[0068] In a specific implementation of the present invention, compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal includes: dynamically correcting the audio signal to be compensated based on the second compensation parameter combined with a filter coefficient to obtain a target audio signal.

[0069] Specifically, the audio signal to be compensated is dynamically corrected based on the second compensation parameter combined with the filter coefficient to obtain a target audio signal, and the target audio signal is transmitted to a speaker, which plays the target audio signal. By correcting the audio signal to be compensated by the second compensation parameter, not only can the played audio be more in line with the user's daily habits, the user no longer needs to adjust the played audio multiple times, which improves the convenience of listening to audio, and can avoid the interference of the spatial environment on the audio signal to the greatest extent. Through artificial intelligence technology, the automatic compensation of the audio signal is realized, and the intelligent control of the audio signal is better realized, so that the final played audio can be closer to the ideal playback effect, improve the user's auditory experience, and make the audio signal compensation achieve a more ideal effect.

[0070] In the embodiment of the present invention, the audio category of the audio signal to be compensated is identified based on the spectrum fluctuation and audio feature information, which can improve the accuracy of audio category identification and greatly reduce the investment of manpower cost. The first compensation parameter of the audio signal to be compensated is determined by equalizer compensation analysis based on user habit information. Taking into account the influence of user habit information on audio signal compensation, the intelligence of audio signal compensation is improved, so that the final played audio is more in line with the user's daily habits. The frequency response data of the audio signal to be compensated is obtained, and the target correction coefficient is generated by impulse response analysis based on the frequency response data combined with the environmental frequency response parameters to adjust the first compensation parameter and generate the second compensation parameter. By analyzing the frequency response data of the audio signal to be compensated, the audio output can be made flatter. At the same time, the final compensation parameter is generated by impulse response analysis combined with the environmental frequency response parameters, which can avoid the interference of the spatial environment on the audio signal to the greatest extent. The audio signal to be compensated is compensated by the second compensation parameter to generate the target audio signal, so as to better realize the intelligent control of the audio signal, so that the final played audio can be closer to the ideal playback effect, improve the user's auditory experience, and achieve a more ideal effect of audio signal compensation.

[0071] Embodiment 2

[0072] See also Figure 2 , Figure 2 : is a flow chart of an audio signal compensation method based on AI control in another embodiment of the present invention, the method is applied to an audio pickup, an embedded real-time operating system and a speaker, the embedded real-time operating system is connected to the audio pickup and the speaker respectively; the method comprises:

[0073] S201: acquiring an audio signal to be compensated based on an audio pickup, and transmitting the audio signal to be compensated to an embedded real-time operating system, wherein the embedded real-time operating system extracts features of the audio signal to be compensated through an artificial intelligence engine to obtain audio feature information;

[0074] S202: Calculating the frequency spectrum fluctuation of the audio signal to be compensated, and performing audio category recognition on the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information to obtain an audio category corresponding to the audio signal to be compensated;

[0075] S203: Acquire user habit information corresponding to the audio category, and determine a first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information;

[0076] S204: performing short-time Fourier transform on the audio signal to be compensated to obtain short-time frequency domain information;

[0077] S205: calculating an autocorrelation function of the audio signal to be compensated based on the short-time frequency domain information, and determining a linear prediction error filter coefficient based on the autocorrelation function;

[0078] In the specific implementation process of the present invention, the autocorrelation function of the audio signal to be compensated is calculated based on the short-time frequency domain information, the square sum of the real part and the imaginary part of each frequency point in the audio signal to be compensated is calculated according to the short-time frequency domain information, and the target envelope of the audio signal to be compensated after segmentation processing is determined. The envelope of the audio refers to controlling various parameters of the audio through a function curve to achieve dynamic changes in volume and other effects. The envelope of the audio signal to be compensated after segmentation processing is flattened according to the target envelope to obtain the audio signal to be compensated after the envelope flattening. The audio signal to be compensated after the envelope flattening is inversely fast Fourier transformed according to the square sum to obtain the autocorrelation function, and the linear prediction error filter coefficient is determined based on the autocorrelation function. The autocorrelation function is solved and calculated using the Levinson-Dubin algorithm. The Levinson-Dubin algorithm is an algorithm for linear prediction and calculation of autocorrelation values ​​to obtain linear prediction coefficients, and the linear prediction error filter coefficients are generated according to the linear prediction coefficients.

[0079] S206: performing frequency response calculation based on the linear prediction error filter coefficient using a sliding window and a random suppression curve to obtain frequency response data;

[0080] In the specific implementation process of the present invention, the frequency response is calculated based on the linear prediction error filter coefficient using a sliding window and a random suppression curve, the enhancement strength of the audio signal is determined by simulating the audio time domain signal after the frequency response, the random suppression curve is determined according to the enhancement strength, the amplitude jump is smoothed by the sliding window, so that the amplitude difference between each frequency point of the audio signal is readjusted, and the frequency response is calculated by combining the linear prediction error filter coefficient with the amplitude jump smoothing data obtained by the sliding window and the enhancement strength obtained by the random suppression curve to obtain frequency response data. The frequency response data captures the relative difference in the amplitude of the sound signal at different frequencies. The true coefficient generated by the frequency response data can better adjust the flatness of the audio signal.

[0081] S207: The embedded real-time operating system controls the audio pickup to collect the test audio signal of the current environment, and obtains the audio parameters of the test audio signal;

[0082] In the specific implementation process of the present invention, the embedded real-time operating system controls each speaker to play the test audio in the current environment in turn, and the embedded real-time operating system controls the audio pickup to collect the test audio signal of the current environment. The test audio signal is generated by the current environment, and the audio parameters of the test audio signal, such as loudness, are obtained.

[0083] S208: Generate environmental frequency response parameters based on audio parameters of the test audio signal combined with audio parameters of a preset standard audio signal;

[0084] In a specific implementation of the present invention, an environmental frequency response parameter is generated based on audio parameters of a test audio signal combined with audio parameters of a preset standard audio signal, wherein the preset standard audio signal is an audio signal without environmental absorption, and the audio parameters of the test audio signal are compared with the audio parameters of the preset standard audio signal. The difference frequency band between the two frequency response parameters is determined according to the comparison result, and the environmental frequency response parameter is generated according to the difference frequency band. The influence of the current environment on the audio signal playback can be corrected through the environmental frequency response parameters.

[0085] S209: Acquire the peak occurrence time of the test audio signal, and perform impulse response analysis based on the peak occurrence time in combination with a preset head-related transfer function to obtain impulse response data;

[0086] In the specific implementation process of the present invention, the peak occurrence time of the test audio signal is obtained, and an impulse response analysis is performed based on the peak occurrence time in combination with a preset head-related transfer function. A preset user model is extracted in the simulation software, and the user model includes a head model and a torso model. The finite difference method is used to perform multiple scattering calculations using the user model in combination with the Green's function to obtain the head-related transfer function, and the first impulse response data in the high-frequency band of the head-related transfer function is obtained. The peak occurrence time of each test audio signal is used for calculation through the truncation method to obtain the second impulse response data in the current environment, and the final impulse response data is generated from the first impulse response data and the second impulse response data.

[0087] S210: Calculating based on the frequency response data, the environmental frequency response parameters and the impulse response data to obtain a target correction coefficient;

[0088] S211: adjusting the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter;

[0089] S212: Compensate the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal, and transmit the target audio signal to a speaker, so that the speaker plays the target audio signal.

[0090] In the embodiment of the present invention, the audio category of the audio signal to be compensated is identified based on the spectrum fluctuation and audio feature information, which can improve the accuracy of audio category identification and greatly reduce the investment of manpower cost. The first compensation parameter of the audio signal to be compensated is determined by equalizer compensation analysis based on user habit information. Taking into account the influence of user habit information on audio signal compensation, the intelligence of audio signal compensation is improved, so that the final played audio is more in line with the user's daily habits. The frequency response data of the audio signal to be compensated is obtained, and the target correction coefficient is generated by impulse response analysis based on the frequency response data combined with the environmental frequency response parameters to adjust the first compensation parameter and generate the second compensation parameter. By analyzing the frequency response data of the audio signal to be compensated, the audio output can be made flatter. At the same time, the final compensation parameter is generated by impulse response analysis combined with the environmental frequency response parameters, which can avoid the interference of the spatial environment on the audio signal to the greatest extent. The audio signal to be compensated is compensated by the second compensation parameter to generate the target audio signal, so as to better realize the intelligent control of the audio signal, so that the final played audio can be closer to the ideal playback effect, improve the user's auditory experience, and achieve a more ideal effect of audio signal compensation.

[0091] Embodiment 3

[0092] See also Figure 3 , Figure 3 1 is a schematic diagram of the structural composition of an audio signal compensation system based on AI control in an embodiment of the present invention, wherein the system comprises an audio pickup 31, an embedded real-time operating system 32 and a speaker 33, wherein the embedded real-time operating system 32 is connected to the audio pickup 31 and the speaker 33 respectively; the system is configured to execute the audio signal compensation method based on AI control described in the above embodiment.

[0093] In the specific implementation process of the present invention, the audio pickup 31 is used to obtain the audio signal to be compensated, and the embedded real-time operating system 32 has a built-in AI engine. The AI ​​engine automatically performs relevant operations for audio signal compensation, can control the work of various components of the system, and realize AI control of audio signal compensation, so that the system can make corresponding decisions independently, and the speaker 33 is used to play audio. For example, it is used for optimizing the audio during the live broadcast process and optimizing the audio by the music playback software. Intelligent compensation of audio signals is realized through artificial intelligence technology, and the entire audio signal compensation processing process is realized through AI control, which greatly improves the intelligence and convenience of audio signal compensation. Figure 3 The system shown does not constitute a limitation of all components, and may include more or less components than shown, or combine certain components.

[0094] In the specific implementation process of the present invention, the specific implementation methods of the system items can be referred to the above embodiments, which will not be repeated here.

[0095] In the embodiment of the present invention, the audio category of the audio signal to be compensated is identified based on the spectrum fluctuation and audio feature information, which can improve the accuracy of audio category identification and greatly reduce the investment of manpower cost. The first compensation parameter of the audio signal to be compensated is determined by equalizer compensation analysis based on user habit information. Taking into account the influence of user habit information on audio signal compensation, the intelligence of audio signal compensation is improved, so that the final played audio is more in line with the user's daily habits. The frequency response data of the audio signal to be compensated is obtained, and the target correction coefficient is generated by impulse response analysis based on the frequency response data combined with the environmental frequency response parameters to adjust the first compensation parameter and generate the second compensation parameter. By analyzing the frequency response data of the audio signal to be compensated, the audio output can be made flatter. At the same time, the final compensation parameter is generated by impulse response analysis combined with the environmental frequency response parameters, which can avoid the interference of the spatial environment on the audio signal to the greatest extent. The audio signal to be compensated is compensated by the second compensation parameter to generate the target audio signal, so as to better realize the intelligent control of the audio signal, so that the final played audio can be closer to the ideal playback effect, improve the user's auditory experience, and achieve a more ideal effect of audio signal compensation.

[0096] Embodiment 4

[0097] See also Figure 4 , Figure 4 : is a structural schematic diagram of an audio signal compensation device based on AI control in an embodiment of the present invention, the device is applied to an audio pickup, an embedded real-time operating system and a speaker, the embedded real-time operating system is connected to the audio pickup and the speaker respectively; the device includes:

[0098] The audio feature extraction module 41 is used to obtain the audio signal to be compensated based on the audio pickup, and transmit the audio signal to be compensated to the embedded real-time operating system, and the embedded real-time operating system extracts the features of the audio signal to be compensated through the artificial intelligence engine to obtain audio feature information;

[0099] The audio category recognition module 42 is used to calculate the frequency spectrum fluctuation of the audio signal to be compensated, and perform audio category recognition on the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated;

[0100] A first compensation parameter acquisition module 43 is used to acquire user habit information corresponding to the audio category, and determine a first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information;

[0101] The second compensation parameter acquisition module 44 is used to acquire the frequency response data of the audio signal to be compensated, generate a target correction coefficient based on the frequency response data combined with the environmental frequency response parameters by using impulse response analysis, and adjust the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter;

[0102] The audio signal compensation module 45 is configured to compensate the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal, and transmit the target audio signal to a speaker, so that the speaker plays the target audio signal.

[0103] In the specific implementation process of the present invention, the specific implementation methods of the device items can be referred to the above embodiments, which will not be repeated here.

[0104] In the embodiment of the present invention, the audio category of the audio signal to be compensated is identified based on the spectrum fluctuation and audio feature information, which can improve the accuracy of audio category identification and greatly reduce the investment of manpower cost. The first compensation parameter of the audio signal to be compensated is determined by equalizer compensation analysis based on user habit information. Taking into account the influence of user habit information on audio signal compensation, the intelligence of audio signal compensation is improved, so that the final played audio is more in line with the user's daily habits. The frequency response data of the audio signal to be compensated is obtained, and the target correction coefficient is generated by impulse response analysis based on the frequency response data combined with the environmental frequency response parameters to adjust the first compensation parameter and generate the second compensation parameter. By analyzing the frequency response data of the audio signal to be compensated, the audio output can be made flatter. At the same time, the final compensation parameter is generated by impulse response analysis combined with the environmental frequency response parameters, which can avoid the interference of the spatial environment on the audio signal to the greatest extent. The audio signal to be compensated is compensated by the second compensation parameter to generate the target audio signal, so as to better realize the intelligent control of the audio signal, so that the final played audio can be closer to the ideal playback effect, improve the user's auditory experience, and achieve a more ideal effect of audio signal compensation.

[0105] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the audio signal compensation method based on AI control of any one of the above embodiments is implemented. The computer-readable storage medium includes but is not limited to any type of disk (including floppy disk, hard disk, optical disk, CD-ROM, and magneto-optical disk), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic card or optical card. That is, the storage device includes any medium that can store or transmit information in a readable form by a device (for example, a computer, a mobile phone), which can be a read-only memory, a disk or an optical disk, etc.

[0106] In addition, the above is a detailed introduction to an audio signal compensation method based on AI control and related devices provided in an embodiment of the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for a general technician in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An audio signal compensation method based on AI control, characterized in that: The method is applied to an audio pickup, an embedded real-time operating system and a loudspeaker, wherein the embedded real-time operating system is connected to the audio pickup and the loudspeaker respectively; the method comprises: Acquire an audio signal to be compensated based on an audio pickup, and transmit the audio signal to be compensated to an embedded real-time operating system, wherein the embedded real-time operating system extracts features of the audio signal to be compensated through an artificial intelligence engine to obtain audio feature information; Calculating the frequency spectrum fluctuation of the audio signal to be compensated, and performing audio category recognition on the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information to obtain an audio category corresponding to the audio signal to be compensated; Acquire user habit information corresponding to the audio category, and determine a first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information; Acquire frequency response data of the audio signal to be compensated, generate a target correction coefficient by using impulse response analysis based on the frequency response data combined with an environmental frequency response parameter, and adjust the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter; Compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal, and transmitting the target audio signal to a speaker, so that the speaker plays the target audio signal; The method of calculating the spectrum fluctuation of the audio signal to be compensated, and identifying the audio category of the audio signal to be compensated based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated includes: obtaining the high-frequency band kurtosis of the spectrum of each audio frame in the audio signal to be compensated; obtaining the spectrum of each audio frame in the audio signal to be compensated, and calculating the logarithmic energy difference of the spectrum of each audio frame in the mid- and low-frequency band spectrum; obtaining the Mel-frequency cepstral coefficients of each audio frame in the audio signal to be compensated, performing natural logarithm calculation processing based on the Mel-frequency cepstral coefficients to obtain a logarithmic energy vector, and generating energy fluctuation data based on the logarithmic energy vector; generating spectrum fluctuation based on the high-frequency band kurtosis of the spectrum, the logarithmic energy difference and the energy fluctuation data; inputting the spectrum fluctuation and audio feature information into an audio classification model, and the audio classification model identifying the audio category based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated, wherein the audio classification model includes a residual gated convolutional neural network based on a channel attention mechanism and a bidirectional long short-term memory network; The method of determining the first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information includes: determining a corresponding compensation filter based on the audio category, and obtaining a first gain coefficient of the compensation filter based on the user habit information; calculating a timbre gain factor and a volume gain factor based on the user habit information combined with the audio feature information; obtaining a frequency distribution curve of the audio signal to be compensated, determining the amplitude value of each frequency point in the frequency distribution curve, and determining a second gain coefficient based on the user habit information combined with the amplitude value of each frequency point; calculating an equalizer gain coefficient based on the first gain coefficient and the second gain coefficient, and determining the first compensation parameter of the audio signal to be compensated based on the equalizer gain coefficient, the timbre gain factor and the volume gain factor.

2. The audio signal compensation method based on AI control according to claim 1, characterized in that: The extracting features of the audio signal to be compensated to obtain audio feature information includes: Preprocessing the audio signal to be compensated to obtain a preprocessed audio signal to be compensated; Performing segmentation processing on the preprocessed audio signal to be compensated to obtain the segmented audio signal to be compensated; Extracting short-time audio features from the audio signal to be compensated after segmentation to obtain short-time audio feature information; The long-term audio feature is extracted from the audio signal to be compensated after the segmentation process to obtain long-term audio feature information, and audio feature information is generated based on the short-term audio feature information and the long-term audio feature information.

3. The audio signal compensation method based on AI control according to claim 1, characterized in that: The obtaining the frequency response data of the audio signal to be compensated comprises: Performing short-time Fourier transform on the audio signal to be compensated to obtain short-time frequency domain information; Calculating an autocorrelation function of the audio signal to be compensated based on the short-time frequency domain information, and determining a linear prediction error filter coefficient based on the autocorrelation function; Frequency response calculation is performed based on the linear prediction error filter coefficient using a sliding window and a random suppression curve to obtain frequency response data.

4. The audio signal compensation method based on AI control according to claim 1, characterized in that: The generating a target correction coefficient by using impulse response analysis based on the frequency response data combined with the environmental frequency response parameters includes: The embedded real-time operating system controls the audio pickup to collect the test audio signal of the current environment, and obtains the audio parameters of the test audio signal; Generate environmental frequency response parameters based on audio parameters of the test audio signal combined with audio parameters of a preset standard audio signal; Acquire the peak occurrence time of the test audio signal, perform impulse response analysis based on the peak occurrence time in combination with a preset head-related transfer function, and obtain impulse response data; A target correction coefficient is obtained by performing calculation based on the frequency response data, the environmental frequency response parameters and the impulse response data.

5. The audio signal compensation method based on AI control according to claim 1, characterized in that: The compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal includes: The audio signal to be compensated is dynamically corrected based on the second compensation parameter in combination with the filter coefficient to obtain a target audio signal.

6. An audio signal compensation device based on AI control, characterized in that: Applied to an audio pickup, an embedded real-time operating system and a loudspeaker, wherein the embedded real-time operating system is connected to the audio pickup and the loudspeaker respectively; the device comprises: Audio feature extraction module: used to obtain the audio signal to be compensated based on the audio pickup, and transmit the audio signal to be compensated to the embedded real-time operating system, and the embedded real-time operating system extracts the features of the audio signal to be compensated through the artificial intelligence engine to obtain audio feature information; An audio category recognition module is used to calculate the frequency spectrum fluctuation of the audio signal to be compensated, and to identify the audio category of the audio signal to be compensated based on the frequency spectrum fluctuation and audio feature information, so as to obtain the audio category corresponding to the audio signal to be compensated; A first compensation parameter acquisition module: used to acquire user habit information corresponding to the audio category, and determine the first compensation parameter of the audio signal to be compensated by using equalizer compensation analysis based on the user habit information; A second compensation parameter acquisition module is used to acquire frequency response data of the audio signal to be compensated, generate a target correction coefficient based on the frequency response data combined with the ambient frequency response parameters by using impulse response analysis, and adjust the first compensation parameter based on the target correction coefficient to obtain a second compensation parameter; An audio signal compensation module: used for compensating the audio signal to be compensated based on the second compensation parameter to obtain a target audio signal, and transmitting the target audio signal to a speaker, and the speaker plays the target audio signal; The method of calculating the spectrum fluctuation of the audio signal to be compensated, and identifying the audio category of the audio signal to be compensated based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated includes: obtaining the high-frequency band kurtosis of the spectrum of each audio frame in the audio signal to be compensated; obtaining the spectrum of each audio frame in the audio signal to be compensated, and calculating the logarithmic energy difference of the spectrum of each audio frame in the mid- and low-frequency band spectrum; obtaining the Mel-frequency cepstral coefficients of each audio frame in the audio signal to be compensated, performing natural logarithm calculation processing based on the Mel-frequency cepstral coefficients to obtain a logarithmic energy vector, and generating energy fluctuation data based on the logarithmic energy vector; generating spectrum fluctuation based on the high-frequency band kurtosis of the spectrum, the logarithmic energy difference and the energy fluctuation data; inputting the spectrum fluctuation and audio feature information into an audio classification model, and the audio classification model identifying the audio category based on the spectrum fluctuation and audio feature information to obtain the audio category corresponding to the audio signal to be compensated, wherein the audio classification model includes a residual gated convolutional neural network based on a channel attention mechanism and a bidirectional long short-term memory network; The method of determining the first compensation parameter of the audio signal to be compensated by using an equalizer compensation analysis based on the user habit information includes: determining a corresponding compensation filter based on the audio category, and obtaining a first gain coefficient of the compensation filter based on the user habit information; calculating a timbre gain factor and a volume gain factor based on the user habit information combined with the audio feature information; obtaining a frequency distribution curve of the audio signal to be compensated, determining the amplitude value of each frequency point in the frequency distribution curve, and determining a second gain coefficient based on the user habit information combined with the amplitude value of each frequency point; calculating an equalizer gain coefficient based on the first gain coefficient and the second gain coefficient, and determining the first compensation parameter of the audio signal to be compensated based on the equalizer gain coefficient, the timbre gain factor and the volume gain factor.

7. An audio signal compensation system based on AI control, characterized in that: The system includes an audio pickup, an embedded real-time operating system and a speaker, wherein the embedded real-time operating system is connected to the audio pickup and the speaker respectively, and the system is configured to execute the audio signal compensation method based on AI control as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on an electronic device, the electronic device executes the audio signal compensation method based on AI control as described in any one of claims 1 to 5.

Images