A method for extracting acoustic signal features of a gas insulated device

Abstract

The application discloses a kind of acoustic signal feature extraction methods of gas insulated equipment, comprising: the sound signal collected is carried out noise reduction processing, and the mean and variance of signal are normalized processing;Using short-time Fourier transform STFT and wavelet transform WT, the sound signal of pre-processing is carried out time-frequency decomposition;Comprehensive MFCC, AWPC adaptive wavelet packet coefficient and instantaneous frequency feature IF are carried out comprehensive extraction to the feature of sound signal;The voiceprint feature extracted is carried out feature fusion, and MFCC, AWPC and IF feature are spliced, form a high-dimensional feature vector, then using principal component analysis PCA is carried out dimension reduction;The feature after dimension reduction is standardized, and correlation analysis method is further used to screen features.The application can extract effective voiceprint features from the sound signal generated by gas insulated equipment defects and reduce the influence of noise on detection, thereby improving the accuracy of detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of gas insulated defect acoustic signal feature extraction, and particularly relates to a gas insulated equipment acoustic signal feature extraction method. BACKGROUND

[0002] In the field of gas insulation defect detection, accurately extracting the acoustic signal features generated by partial discharge and mechanical vibration plays an extremely important role in determining the defect type. Although the traditional feature extraction method, such as the Mel frequency cepstral coefficient (MFCC), can capture the frequency domain information of the sound signal to a certain extent, it still has certain limitations in processing noise, sound signal changes and detailed features. In order to improve the recognition accuracy and robustness, the present application proposes a feature extraction method that fuses time-frequency and frequency domain information. Time-frequency analysis methods such as short-time Fourier transform (STFT) and wavelet transform (WT) can provide signal information at different scales and time resolutions. However, these methods alone still cannot fully extract all the features in the sound signal. Therefore, the present application proposes a time-frequency domain feature fusion (STFF) algorithm that combines MFCC, AWPC (adaptive wavelet packet coefficient) and instantaneous frequency feature (IF) to comprehensively capture the subtle features and dynamic changes of the acoustic signal, thereby improving the accuracy and robustness of the acoustic signal recognition system in the extraction of acoustic signal features of gas insulated equipment. SUMMARY

[0003] In view of the above-mentioned existing problems, the present application is proposed.

[0004] According to a first aspect of the present application, a gas insulated equipment acoustic signal feature extraction method is provided, which can extract effective acoustic signal features from the sound signal generated by the defect of the gas insulated equipment and reduce the influence of noise on detection, thereby improving the detection accuracy.

[0005] To solve the above technical problems, the present application provides the following technical solution, a gas insulated equipment acoustic signal feature extraction method, comprising:

[0006] The collected sound signal is subjected to noise reduction processing, and the mean and variance of the signal are subjected to normalization processing. The preprocessed sound signal is subjected to time-frequency decomposition by using short-time Fourier transform (STFT) and wavelet transform (WT). The features of the sound signal are comprehensively extracted by using MFCC, AWPC (adaptive wavelet packet coefficient) and IF (instantaneous frequency feature). The extracted acoustic signal features are fused, the MFCC, AWPC and IF features are spliced to form a high-dimensional feature vector, and the principal component analysis (PCA) is used for dimension reduction. The dimension-reduced features are subjected to standardization processing to improve the stability of the algorithm, and the correlation analysis method is used for further feature screening.

[0007] As a preferred scheme of the acoustic signal feature extraction method of the gas insulated equipment, the noise reduction processing comprises: adopting a Wiener filter to perform noise reduction processing on the sound signal, assuming that an observation signal x(t) satisfies the following relationship:

[0008] x(t) = s(t) + n(t)

[0009] where s(t) is a detection signal, and n(t) is noise;

[0010] The output of the Wiener filter is The output of the Wiener filter is obtained by convolution:

[0011]

[0012] The objective of the Wiener filter is to minimize the mean square error E:

[0013]

[0014] In the frequency domain, assuming that the power spectral densities of the signal and the noise are S sx (f) and S n (f) respectively, the frequency response H(f) of the filter is calculated by the following formula:

[0015]

[0016] In the formula, S sx (f) is the cross power spectral density of the signal and the observation signal; S x (f) = S sx (f) + S n (f is the power spectral density of the observation signal.

[0017] As a preferred scheme of the acoustic signal feature extraction method of the gas insulated equipment, the noise reduction processing comprises: if the power spectral densities of the signal s(t) and the noise n(t) are known, performing inverse Fourier transform to the time domain to obtain a time domain filter h(t).

[0018] The mean value of the signal after the filtering processing is calculated, and the mean value of each frame of signal is subtracted, so that the mean value of the signal in each frame is zero, the influence of the DC component on the feature is eliminated, and the standard deviation of each frame of signal is adjusted to be a unit, so that the amplitude range of the signal is scaled.

[0019] As a preferred scheme of the acoustic signal feature extraction method of the gas insulated equipment, the time-frequency decomposition comprises: adopting a short-time Fourier transform (STFT) and a wavelet transform (WT) to perform time-frequency decomposition on the preprocessed sound signal.

[0020] Short-time Fourier transform (STFT) is a method of analyzing time-frequency characteristics of a signal by applying a time window to the signal and calculating the Fourier transform. For a preprocessed sound signal x[n], the STFT is represented as:

[0021]

[0022] where w[m] is a window function, and a Hamming window is selected;

[0023]

[0024] where m represents a time offset, and ω represents a frequency variable, and the result X(m, ω) of the STFT is a complex number, representing a signal component at time m and frequency ω.

[0025] As a preferred scheme of the gas insulated device acoustic signal feature extraction method, the comprehensive extraction of the features of the sound signal includes capturing time-frequency characteristics of the signal by using adaptive wavelet packet decomposition, performing multi-level wavelet packet decomposition on the signal, recursively decomposing a frequency band of the signal, decomposing each frequency band of the signal into a low-frequency sub-band and a high-frequency sub-band at each level, selecting a decomposition path that best represents the features of the signal, and extracting wavelet packet coefficients from the selected optimal decomposition tree as the features, according to the time and frequency distribution of the signal.

[0026] As a preferred scheme of the gas insulated device acoustic signal feature extraction method, the comprehensive extraction of the features of the sound signal includes extracting instantaneous frequency features of the sound signal, performing Hilbert transform on the preprocessed signal x(t) to obtain an analytic signal z(t):

[0027]

[0028] where is the Hilbert transform of x(t);

[0029] The instantaneous frequency f(t) is the time derivative of the phase of the analytic signal z(t):

[0030]

[0031] The rate of change of the instantaneous frequency can provide dynamic information about the local frequency change of the signal.

[0032] As a preferred scheme of the gas insulated device acoustic signal feature extraction method, the feature fusion of the extracted voiceprint features includes obtaining a comprehensive feature description of the sound signal by combining MFCC, AWPC, and IF:

[0033] MFCC provides the static spectral features of the signal in the mel frequency scale, which is suitable for describing the harmonic structure of the sound signal; AWPC provides the multi-resolution time-frequency features of the signal, which captures the energy distribution and changes of the signal in different frequency bands; IF provides the dynamic change information of the instantaneous frequency of the signal, which is used to analyze the non-stationary signal, including the modulation and jitter characteristics in the sound;

[0034] The feature vectors of MFCC, AWPC and IF are connected, principal component analysis is used for dimension reduction, and the fused feature vector is used for machine learning model classification and recognition.

[0035] As a preferred scheme of the gas insulated device acoustic signal feature extraction method, the standardized processing of the dimension-reduced features includes standardized processing of the dimension-reduced features. The zero-mean unit-variance standardization Z-Score is adopted, the mean value of the features is adjusted to 0, and the standard deviation is adjusted to 1, so that the standardized features conform to the standard normal distribution, the mean value is 0, and the standard deviation is 1;

[0036] Given a feature vector x = [x1, x2, x3…x n ], the mean value μ and the standard deviation σ are respectively:

[0037]

[0038] The feature vector z = [z1, z2, z3…z n ] after Z-Score standardization is calculated as follows:

[0039]

[0040] The feature data is mapped to the same scale to eliminate the dimensional difference of different features;

[0041] The correlation analysis method is used to evaluate the correlation between the features and the target variables after the standardized features, and the Pearson correlation coefficient is used, the value range is [-1, 1]; when the correlation coefficient is close to 1, it indicates that there is a strong positive correlation between the two variables; when the correlation coefficient is close to-1, it indicates that there is a strong negative correlation; when the correlation coefficient is close to 0, it indicates that there is no linear correlation;

[0042] Given two variables X and Y, the calculation formula of the Pearson correlation coefficient r is:

[0043]

[0044] Where and are the mean values of X and Y, respectively.

[0045] According to a second aspect of the present application, a computer device is provided, comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the method for extracting acoustic signal features of a gas insulated device when executing the computer program.

[0046] According to a third aspect of the present application, a computer readable storage medium is provided, which stores a computer program, and wherein the computer program implements the steps of the method for extracting acoustic signal features of a gas insulated device when executed by a processor.

[0047] The method for extracting acoustic signal features of a gas insulated device according to the present application can extract effective voiceprint features from sound signals generated by defects of a gas insulated device and reduce the influence of noise on detection, thereby improving the accuracy of detection. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0049] Figure 1 A flowchart of a method for extracting acoustic signal features of a gas insulated device according to an embodiment of the present application is shown.

[0050] Figure 2 A flowchart of a method for extracting acoustic signal features of a gas insulated device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned objects, features and advantages of the present application more apparent and understandable, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0052] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced without the specific details, which are not described herein. In other instances, well-known methods, procedures and components have not been described in detail so as not to obscure the present application. Accordingly, the present application is not limited to the embodiments described herein, but rather the scope of the application is to be decided by the appended claims.

[0053] Secondly, the "one embodiment" or "an embodiment" referred to herein means a specific feature, structure, or characteristic in at least one implementation of the application. The appearances of "in one embodiment" or "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive.

[0054] The application is described in detail in conjunction with the schematic drawings. In the detailed description of the embodiments of the application, the sectional views of the device structure are partially enlarged without the general scale for the convenience of illustration, and the schematic drawings are only examples, which should not limit the scope of protection of the application. In addition, the three-dimensional spatial dimensions of length, width and depth should be included in the actual manufacture.

[0055] Meanwhile, in the description of the application, it should be noted that the positions or relationships indicated by the terms "upper, lower, inner and outer" are based on the positions or relationships shown in the drawings, and are only for the convenience of describing the application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the application. In addition, the terms "first, second or third" are only for the purpose of description, and cannot be understood as indicating or implying relative importance.

[0056] Unless otherwise explicitly specified and limited, the terms "mounting, connecting, connecting" in the application should be understood broadly, for example: it can be fixed connection, detachable connection or integral connection; it can also be mechanical connection, electrical connection or direct connection, it can also be indirectly connected through intermediate medium, or it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the application can be understood according to the specific circumstances.

[0057] Embodiment 1, refer to Figure 1 As the first embodiment of the application, the embodiment provides a gas insulated device acoustic signal feature extraction method, comprising:

[0058] S1: The collected sound signal is subjected to noise reduction processing, and the mean and variance of the signal are subjected to normalization processing.

[0059] S2: Short-time Fourier transform (STFT) and wavelet transform (WT) are used to perform time-frequency decomposition on the preprocessed sound signal.

[0060] S3: MFCC, AWPC (adaptive wavelet packet coefficient) and instantaneous frequency feature (IF) are comprehensively extracted for the features of the sound signal, wherein the extraction of MFCC parameters refers to Figure 2 .

[0061] S4: The extracted voiceprint features are fused, the MFCC, AWPC and IF features are spliced to form a high-dimensional feature vector, and principal component analysis (PCA) is used for dimension reduction to reduce the feature dimension and retain the most important information.

[0062] S5: The dimension-reduced features are standardized to improve the stability of the algorithm, and the correlation analysis method is used to further screen the most recognizable features.

[0063] The specific method of step S1 is:

[0064] The Wiener filter method is used for noise reduction processing of the sound signal, assuming that the observed signal x(t) satisfies the following relationship:

[0065] x(t) = s(t) + n(t) (1)

[0066] Where s(t) is the detected signal, and n(t) is the noise.

[0067] The output of the Wiener filter is By convolution, we get:

[0068]

[0069] The goal of the Wiener filter is to minimize the mean square error E:

[0070]

[0071] In the frequency domain, assuming that the power spectral densities of the signal and the noise are S sx (f) and S n (f), respectively, the frequency response H(f) of the filter can be calculated by the following formula:

[0072]

[0073] In the formula, S sx (f) is the cross-power spectral density of the signal and the observed signal; S x (f) = S sx (f) + S n (f) is the power spectral density of the observed signal.

[0074] If the power spectral densities of the signal s(t) and the noise n(t) are known, the frequency domain response of the Wiener filter is formula (4), and its inverse Fourier transform to the time domain can obtain the time domain filter h(t).

[0075] The mean value of the filtered signal is calculated, and the mean value of each frame of signal is subtracted to ensure that the mean value of the signal within each frame is zero. This can eliminate the influence of the DC component on the feature. Then, the standard deviation of each frame of signal is adjusted to be a unit to scale the amplitude range of the signal. It is intended to avoid the instability of the signal feature caused by the change of signal intensity.

[0076] The specific method of the step S2 is:

[0077] The short-time Fourier transform (STFT) and wavelet transform (WT) are used for time-frequency decomposition of the preprocessed sound signal.

[0078] The short-time Fourier transform (STFT) is a method for analyzing the time-frequency characteristics of a signal by applying a time window to the signal and calculating the Fourier transform. For the preprocessed sound signal x[n], the STFT can be expressed as:

[0079]

[0080] where w[m] is a window function, and Hamming is selected.

[0081]

[0082] m represents the time offset, and ω represents the frequency variable. The result X(m, ω) of the STFT is a complex number, representing the signal component at time m and frequency ω.

[0083] By selecting appropriate window length and overlap rate, the STFT can achieve a good balance between time and frequency resolution. The selection of window function and the setting of window length have a direct impact on frequency resolution and time resolution. Short window provides better time resolution, but poor frequency resolution; long window is the opposite. Different window settings can be selected for different defect types to achieve the best extraction accuracy.

[0084] In order to further analyze the non-stationary characteristics of the sound signal, the wavelet transform (WT) is also used to process and analyze the sound signal. The wavelet transform can provide detailed information of the signal at different time and frequency scales through the use of scaled and translated wavelet basis functions for multi-resolution analysis of the signal.

[0085] The specific method of the step S3 is:

[0086] The MFCC, AWPC and IF methods are combined to comprehensively extract the features of the sound signal.

[0087] First, the features of the sound signal are extracted by MFCC, and the steps of MFCC extraction are as follows Figure 1 as shown.

[0088] Secondly, adaptive wavelet packet decomposition is used to capture the time-frequency features of the signal. Wavelet packet decomposition is an extension of wavelet decomposition that can refine the frequency bands of a signal to a higher resolution. The signal is decomposed by multiple levels of wavelet packet decomposition, recursively decomposing the frequency bands of the signal. Each level of decomposition further decomposes each frequency band of the signal into low and high frequency subbands. Wavelet packet decomposition is more flexible than traditional wavelet decomposition because it allows decomposition of arbitrary frequency bands rather than fixed low and high frequencies. The decomposition path that best represents the features of the signal is selected, which can reduce unnecessary computation and preserve the significant features of the signal. The wavelet packet coefficients from the selected optimal decomposition tree are extracted as features. AWPC can provide rich time-frequency features based on the time and frequency distribution of the signal.

[0089] Finally, the instantaneous frequency feature of the sound signal is extracted. The Hilbert transform is performed on the preprocessed signal x(t) to obtain its analytic signal z(t).

[0090]

[0091] where is the Hilbert transform of x(t).

[0092] The instantaneous frequency f(t) is the time derivative of the phase of the analytic signal z(t):

[0093]

[0094] The rate of change of the instantaneous frequency can provide dynamic information about the local frequency change of the signal and is an important tool for non-stationary signal analysis.

[0095] The specific method of the step S4 is:

[0096] By combining MFCC, AWPC and IF, a comprehensive feature description of the sound signal can be obtained:

[0097] MFCC provides the static spectral features of the signal in the Mel frequency scale, which is suitable for describing the harmonic structure of the sound signal.

[0098] AWPC provides the multi-resolution time-frequency features of the signal, which can capture the energy distribution and change of the signal in different frequency bands, and is suitable for various sound signal analysis.

[0099] IF provides dynamic change information of the instantaneous frequency of the signal, which is suitable for analyzing non-stationary signals such as modulation, jitter and other features in sound.

[0100] The feature vectors of MFCC, AWPC and IF are connected, and principal component analysis is used for dimension reduction to reduce the feature dimension and retain the most important information. The fused feature vector can be used for machine learning model classification and identification.

[0101] The principal component analysis mainly includes the following steps:

[0102] 1) Data standardization

[0103] 2) Calculate the covariance matrix

[0104] 3) Calculate the eigenvalues and eigenvectors of the covariance matrix

[0105] 4) Select the main eigenvector

[0106] 5) Transform to new feature space

[0107] By projecting high-dimensional features into a low-dimensional subspace, PCA can reduce redundant information, improve computational efficiency, and enhance model performance.

[0108] The specific method of step S5 is:

[0109] Standardize the dimension-reduced features. Zero-mean unit-variance standardization (Z-Score) is used. Z-Score standardization is the most commonly used standardization method, which adjusts the mean of the feature to 0 and the standard deviation to 1, so that the standardized feature follows the standard normal distribution (mean 0, standard deviation 1).

[0110] Given a feature vector x = [x1, x2, x3…x n ], its mean μ and standard deviation σ are respectively:

[0111]

[0112] The Z-Score standardized feature vector z = [z1, z2, z3…z n ] is calculated as follows:

[0113]

[0114] Mapping feature data to the same scale can eliminate the dimensional differences of different features, make features have the same weight and importance, and improve the convergence speed and stability of the model.

[0115] After the normalized features, the correlation analysis method is used to evaluate the correlation between the features and the target variable. The Pearson correlation coefficient is used in the application, which has a value range of [-1, 1]. When the correlation coefficient is close to 1, it indicates that there is a strong positive correlation between the two variables; when the correlation coefficient is close to -1, it indicates that there is a strong negative correlation; when the correlation coefficient is close to 0, it indicates that there is no linear correlation.

[0116] Given two variables X and Y, the formula for calculating the Pearson correlation coefficient r is:

[0117]

[0118] Where and are the mean values of X and Y, respectively.

[0119] The strongest positive correlation feature quantity is selected as the extracted feature parameter of the sound signal. The feature parameter extraction method proposed in the application can comprehensively capture the subtle features and dynamic changes of the voiceprint by combining the features of time domain, frequency domain and instantaneous frequency, thereby improving the accuracy and robustness of the voiceprint recognition system in the feature extraction of acoustic signals of gas insulated equipment.

[0120] Embodiment 2, the second embodiment of the application, which is different from the previous embodiment:

[0121] If the function is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the application or the parts that essentially contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the various embodiments of the application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various program code storage media.

[0122] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0123] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0124] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing the steps listed in the flowchart block or blocks.

[0125] While the preferred embodiments of the application have been described, additional variations and modifications can be employed by those skilled in the art. Therefore, the appended claims intend to cover all such modifications and variations as fall within the true spirit and scope of the application.

[0126] It is apparent that a person skilled in the art can make various changes and modifications to the application without departing from the spirit and scope thereof. Therefore, if these modifications and changes fall within the scope of the claims and their equivalents, it is intended to include them in the application.

Claims

1. A method of feature extraction of acoustic signals of a gas insulated apparatus, characterized in that: Comprising, The collected sound signal is denoised, and the mean and variance of the signal are normalized; The denoised sound signal is decomposed in time and frequency by using short-time Fourier transform (STFT) and wavelet transform (WT); The features of the sound signal are comprehensively extracted by using MFCC, AWPC adaptive wavelet packet coefficients and instantaneous frequency (IF); The extracted voiceprint features are fused, the MFCC, AWPC and IF features are spliced to form a high-dimensional feature vector, and principal component analysis (PCA) is used for dimension reduction; The features after dimension reduction are standardized, and the correlation analysis method is used for further feature screening.

2. The method of claim 1, wherein: The denoising further comprises using Wiener filtering method to denoise the sound signal, assuming that the observed signal x(t) satisfies the following relationship: x(t) = s(t) + n(t) Where s(t) is the detected signal, and n(t) is the noise; Output of the wiener filter By convolution we obtain: Where x(t) is the observed signal, and h(t) represents the impulse response of the filter; The objective of the Wiener filter is to minimize the mean square error E: In the frequency domain, assume the power spectral densities of the signal and noise are S sx (f) and S n (f), respectively. The frequency response H(f) of the filter is calculated by the following equation: where S sx (f) is the cross power spectral density of the signal and the observation signal; S x (f) = S sx (f) + S n (f) is the power spectral density of the observation signal.

3. The method of claim 2, wherein: The denoising further comprises, if the power spectral density of the signal s(t) and the noise n(t) is known, the inverse Fourier transform is performed to the time domain to obtain the time domain filter h(t); The mean of the filtered signal is calculated, the mean of each frame of signal is subtracted to ensure that the mean of the signal in each frame is zero, the DC component is eliminated, and the standard deviation of each frame of signal is adjusted to be a unit to scale the amplitude range of the signal.

4. The method of claim 3, wherein: The time-frequency decomposition comprises decomposing the preprocessed sound signal in time and frequency by using short-time Fourier transform (STFT) and wavelet transform (WT); Short-time Fourier transform (STFT) is a method for analyzing the time-frequency characteristics of a signal by applying a time window to the signal and calculating the Fourier transform. For the preprocessed sound signal x(n), STFT is represented as: Where w(n) is a window function, and Hamming window is selected. Where m represents the time offset, ω represents the frequency variable, N is a natural number, and a is a constant. For Hamming window, a is 0.

46. The result X(m, ω) of STFT is a complex number, which represents the signal component at time m and frequency ω.

5. A method of extracting acoustic signal features of a gas insulated device according to claim 4, characterized in that: The comprehensive extraction of the features of the sound signal comprises capturing the time-frequency characteristics of the signal by using adaptive wavelet packet decomposition, decomposing the signal in multiple levels of wavelet packet, recursively decomposing the frequency band of the signal, decomposing each frequency band of the signal into low-frequency and high-frequency subbands at each level of decomposition, selecting the optimal decomposition path representing the features of the signal, and extracting the wavelet packet coefficients from the selected optimal decomposition tree as the features.

6. A method of extracting acoustic signal features of a gas insulated device according to claim 5, characterized in that: The comprehensive extraction of the features of the sound signal further comprises extracting the instantaneous frequency (IF) of the sound signal, performing Hilbert transform on the preprocessed signal x(t) to obtain its analytic signal z(t): wherein is the Hilbert transform of x(t), is the imaginary part; The instantaneous frequency f(t) is the time derivative of the phase of the analytic signal z(t): The rate of change of the instantaneous frequency can provide dynamic information about the local frequency change of the signal.

7. A method of extracting acoustic signal features of a gas insulated device according to claim 6, characterized in that: The feature fusion of the extracted voiceprint features comprises: obtaining comprehensive feature description of the sound signal by combining MFCC, AWPC and IF: MFCC provides static spectral features of the signal in the mel frequency scale, which is suitable for describing the harmonic structure of the sound signal; AWPC provides multi-resolution time-frequency features of the signal, which captures the energy distribution and changes of the signal in different frequency bands; IF provides dynamic change information of the instantaneous frequency of the signal, which is used to analyze non-stationary signals, including modulation and jitter features in the sound; The feature vectors of MFCC, AWPC and IF are connected, and principal component analysis is used for dimension reduction; the machine learning model classifies and identifies the fused feature vectors.

8. The method of claim 7, wherein: The standardization processing of the dimension-reduced features further comprises: Z-Score is used for zero-mean unit-variance standardization, which adjusts the mean value of the features to 0 and the standard deviation to 1, so that the standardized features conform to the standard normal distribution with a mean value of 0 and a standard deviation of 1; Given a feature vector x = [x1, x2, x3…x n ], the mean μ and standard deviation σ are respectively: where n is the eigenvector length, x i is the corresponding component of the eigenvector; Z-Score standardized feature vector z = [z1, z2, z3...z n ] is calculated as follows: The feature data is mapped to the same scale to eliminate the dimensional differences of different features; The correlation analysis method is used to evaluate the correlation between the features and the target variables, and the Pearson correlation coefficient is used, which has a value range of [-1, 1]; When the correlation coefficient is close to 1, it indicates that there is a strong positive correlation between the two variables; when the correlation coefficient is close to -1, it indicates that there is a strong negative correlation; when the correlation coefficient is close to 0, it indicates that there is no linear correlation; Given two variables X and Y, the calculation formula of the Pearson correlation coefficient r is: wherein and are the mean values of X and Y, respectively, X i and Y i are the ith values of X and Y, respectively, i being a natural number. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. The processor executes the computer program to realize the steps of the method of any one of claims 1 to 8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1 to 8.

Images