Intelligent signal processing method and system

Abstract

The present invention discloses an intelligent signal processing method and system, which includes signal acquisition, signal preprocessing, frequency domain analysis, local mean decomposition, mask signal processing, feature extraction, classification verification and signal supervision. The present invention belongs to the field of signal processing, and specifically refers to an intelligent signal processing method and system. This solution introduces a dynamic pre-emphasis coefficient and a window function for signal preprocessing; performs frequency domain analysis based on dynamic filtering adjustment, uses a dynamic filter to adjust the spectrum, removes noise and retains effective signals; improves decomposition efficiency and reduces errors by optimizing boundary conditions; thereby improving the accuracy of signal monitoring; enhances high-order features by cepstrum appreciation, and improves the classification and discrimination of signals; makes feature extraction more targeted and flexible based on the design of context-aware filters; optimizes the signal classification model based on the design of position update factors, position shrinkage functions and evolutionary step sizes, thereby improving the effect of signal monitoring.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to an intelligent signal processing method and system. Background Art

[0002] Signal processing methods essentially analyze, process, and optimize signals through a series of mathematical, algorithmic, and technical means to extract useful information, remove noise, enhance signal quality, or restore the signal. However, common signal processing methods suffer from inadequate noise processing, insufficient differentiation between noise and signal components, and significant deviations in reconstructed signals, leading to low signal monitoring accuracy. They also suffer from inadequate signal screening, failing to accurately remove noise components, and ignoring high-order signal features, resulting in low signal classification accuracy and poor signal monitoring. Summary of the Invention

[0003] In response to the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent signal processing method and system. In view of the problems that general signal processing methods have improper noise processing, insufficient distinction between noise and signal components, and serious deviation in reconstructed signals, which lead to low accuracy of signal monitoring, this scheme introduces dynamic pre-emphasis coefficients and window functions for signal preprocessing; frequency domain analysis is performed based on dynamic filtering adjustment, and a dynamic filter is used to adjust the spectrum to remove noise and retain effective signals; by optimizing boundary conditions, the decomposition efficiency is improved and the error is reduced; thereby improving the accuracy of signal monitoring; in view of the problems that general signal processing methods have improper signal screening and cannot accurately remove noise components; high-order features in the signal are ignored, resulting in low signal classification accuracy and poor signal monitoring effect, this scheme enhances high-order features through cepstrum appreciation, improves the classification discrimination of signals, especially the sensitivity to small changes; based on the design of context-aware filters, feature extraction is made more targeted and flexible; based on the design of position update factors, position shrinkage functions and evolutionary step sizes, the signal classification model is optimized, thereby improving the effect of signal monitoring.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides an intelligent signal processing method, which includes the following steps:

[0005] Step S1: signal acquisition;

[0006] Step S2: signal preprocessing;

[0007] Step S3: frequency domain analysis;

[0008] Step S4: local mean decomposition;

[0009] Step S5: mask signal processing;

[0010] Step S6: feature extraction;

[0011] Step S7: classification verification;

[0012] Step S8: Signal supervision.

[0013] Furthermore, in step S1, the signal acquisition is to acquire a historical electromagnetic signal data set; the historical electromagnetic signal data set includes electromagnetic signal data and signal quality; the signal quality includes high quality, good, average and poor.

[0014] Furthermore, in step S2, the signal preprocessing is to remove noise and enhance the high-frequency characteristics of the electromagnetic signal; the received electromagnetic signal is pre-emphasized, which is expressed as: ; Use the window function to split the electromagnetic signal frame; expressed as: ; ; ;in, It is the electromagnetic signal after pre-emphasis; is the original collected electromagnetic signal; is the dynamic pre-emphasis coefficient; is the basic pre-emphasis coefficient; is the aggravation adjustment factor; is the previous sampling point value of the electromagnetic signal; t is time; n1 is the sampling point index; is the windowing adjustment factor.

[0015] Furthermore, in step S3, the frequency domain analysis is to convert each frame of electromagnetic signal from the time domain to the frequency domain based on the fast Fourier transform, and calculate the amplitude spectrum, which is expressed as: ; Use a triangular filter to map the linear frequency to the Mel frequency scale, expressed as: ; Perform dynamic filter adjustment, expressed as: ; ; The overall filter response is expressed as: ;in, is the amplitude spectrum of the frequency domain signal; N1 is the total number of frequencies; k is the frequency index; is the Fast Fourier Transform; is the mth Mel frequency eigenvalue; log is the logarithmic operation; is the frequency response of the triangular filter; is the filter weight after dynamic adjustment; are the initial filter weights; is the Sigmoid function, x is the function variable; is the input signal; is the signal-to-noise ratio; is the signal-to-noise ratio threshold; is the overall frequency response of the dynamic filter; N2 is the total number of filters; i is the filter index; is the dynamic adjustment weight of the filter at time t; is the frequency response of the i-th filter at the k-th frequency; is the noise power spectral density; is the maximum value of noise; is the filter adjustment coefficient.

[0016] Furthermore, in step S4, the local mean decomposition is the extreme local mean and amplitude calculation, which is expressed as: ; ; The sliding average calculates the local mean and amplitude, which is expressed as: ; ; ; ; The stopping criterion value f is expressed as: ; The preliminary reconstructed signal is expressed as: ;in, It is the local mean of adjacent extreme points that reflects the smooth trend of the electromagnetic reflection signal; It is the local amplitude of adjacent extreme points, reflecting the amplitude change of electromagnetic signal; and are adjacent extreme value points in the electromagnetic signal; It is the result of removing the local mean of the electromagnetic signal; It is the normalized vibration component of the electromagnetic signal; is the local mean in the sliding window; is the local amplitude in the sliding window; is the input signal at time t; k is the length of the sliding window; is the total number of sampling points of the electromagnetic signal; n is the sampling point index of the electromagnetic signal; z is the error signal after the current decomposition; is the nth sampling point value of the signal residual; z1 is the mean value of the signal residual; is the threshold of the stopping criterion; is the signal after preliminary reconstruction; is the eigenmode function; is the decomposed residual signal, j is the total number of IMFs; p is the IMF index; is the amplitude adjustment factor.

[0017] Furthermore, in step S5, the mask signal processing is to calculate the instantaneous frequency using Hilbert transform, which is expressed as: ; Mask signal Expressed as: ; Filter the decomposed components to remove noise and retain the effective electromagnetic signal; the component definition is expressed as: ; The final reconstructed electromagnetic signal Expressed as: ;in, is the instantaneous frequency at time t ; is the phase function of the electromagnetic signal; is the instantaneous amplitude of the component; is the effective modal component after screening; and are the component signals of the positive frequency and negative frequency parts respectively; n3 is the number of effective modal components screened out, and i3 is the corresponding index; is the filtered modal component.

[0018] Furthermore, in step S6, the feature extraction is to take the logarithm of the power spectrum, which is expressed as: ; Perform inverse Fourier transform on the power spectrum logarithm to obtain the cepstral coefficients, which are expressed as: ; Perform cepstral appreciation to enhance the high-order features in the cepstral coefficients, expressed as: ; Introduce context-aware filter, expressed as: ; ; The adjusted cepstral coefficient is expressed as: ; ;in, is the power spectrum of the signal; is the spectrum of the signal; are the cepstral coefficients; is the inverse Fourier transform; are the cepstral coefficients after appreciation; is the original value of the cepstral coefficient; n2 is the order of the cepstral coefficient; L is the length of the rise; is the frequency response of the context-aware filter; is the attenuation coefficient; d is the adjustment parameter; exp(·) is the exponential function; is the adjusted cepstral coefficient; is the basic attenuation coefficient; is the attenuation adjustment factor; f is the current frequency; is the center frequency; It is a random number between 0 and 1, independent each time; is the basic adjustment parameter; is the modulation amplitude control factor.

[0019] Furthermore, in step S7, the classification verification is to classify the processed electromagnetic signal data based on a multi-class SVM, construct a signal classification model, and optimize the signal classification model; specifically including:

[0020] Step S71: Based on the emphasis adjustment factor, adjustment parameter, filter adjustment coefficient, windowing adjustment factor, attenuation adjustment factor, modulation amplitude control factor, amplitude adjustment factor, kernel function and penalty parameter in the multi-class SVM, a parameter optimization space is established, and the individual position in the optimization population is initialized. The classification accuracy obtained based on the individual position is used as the individual fitness value; the individual position initialization is expressed as: ;in, and are the initialization positions of the o+1th individual and the oth individual respectively; p is the control parameter;

[0021] Step S72: Individual location update is expressed as:

[0022] ;

[0023] ;

[0024] ;

[0025] ;

[0026] in, The iteration position of the individual c-th dimension, T is the number of iterations, and Tmax is the maximum number of iterations; is the global optimal solution; is the correction parameter; is the position update factor; is the position shrinkage function; rand is a random number between 0 and 1; is the position of a random individual in the population; is the evolutionary step length; is the percentage difference from the global optimal solution; is a regulating factor; is the minimum weight; is the disturbance factor; is the control factor; is the convergence term; is the maximum development factor; is the variance of the current distribution of the population; is the maximum variance of the initial distribution of the population; is the gradient adjustment factor; is the fitness value gradient;

[0027] Step S73: A fitness threshold is set in advance. When an individual fitness value is higher than the fitness threshold, the optimization is complete, and a signal classification model is established based on the individual position. If the maximum number of iterations is reached, return to step S71; otherwise, continue with position iteration.

[0028] Furthermore, in step S8, the signal supervision is based on the established signal classification model, and electromagnetic signals are collected in real time. After processing in steps S2 to S6, classification processing is performed based on the signal classification model, and signals with average signal quality are marked. When the quality is reduced to poor, an early warning is issued; a warning is issued for signals directly classified as poor, and relevant personnel are notified to conduct inspection and maintenance.

[0029] The present invention provides an intelligent signal processing system, which includes a signal acquisition module, a signal preprocessing module, a frequency domain analysis module, a local mean decomposition module, a mask signal processing module, a feature extraction module, a classification verification module and a signal supervision module;

[0030] The signal acquisition module collects historical electromagnetic signal data sets and sends the data to the signal preprocessing module;

[0031] The signal preprocessing module performs denoising, pre-emphasis and window function segmentation on the signal; and sends the data to the frequency domain analysis module;

[0032] The frequency domain analysis module converts the electromagnetic signal from the time domain to the frequency domain, extracts the frequency characteristics and removes the noise using a dynamic filter; and sends the data to the local mean decomposition module;

[0033] The local mean decomposition module decomposes the signal by optimizing the boundary conditions and the stopping criteria, extracts the smoothing trend and the amplitude change, and sends the data to the mask signal processing module;

[0034] The mask signal processing module calculates the instantaneous frequency and uses mask signal processing to filter the signal components; and sends the data to the feature extraction module;

[0035] The feature extraction module performs logarithmic transformation on the power spectrum of the signal, performs cepstrum analysis and enhances the high-order features of the signal through a context-aware filter; and sends the data to the classification verification module;

[0036] The classification verification module classifies the processed electromagnetic signal data based on a multi-class SVM and adjusts the model through an optimization algorithm; and sends the data to the signal supervision module;

[0037] After real-time signal acquisition, the signal monitoring module processes and classifies the signals based on the classification model, issues early warnings for signals of poor quality, and notifies relevant personnel.

[0038] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0039] (1) In view of the problems that general signal processing methods have inappropriate noise processing, insufficient distinction between noise and signal components, and serious deviation in reconstructed signals, which lead to low accuracy of signal monitoring, this scheme introduces dynamic pre-emphasis coefficients and window functions for signal preprocessing; performs frequency domain analysis based on dynamic filtering adjustment, uses dynamic filters to adjust the spectrum, removes noise and retains valid signals; improves decomposition efficiency and reduces errors by optimizing boundary conditions, thereby improving the accuracy of signal monitoring.

[0040] (2) In view of the problems that general signal processing methods have inappropriate signal screening and cannot accurately remove noise components; they ignore high-order features in the signal, resulting in low signal classification accuracy and poor signal monitoring effect. This scheme enhances high-order features through cepstral enhancement, improves the classification and discrimination of signals, especially the sensitivity to small changes; designs context-aware filters to make feature extraction more targeted and flexible; and optimizes the signal classification model based on the design of position update factors, position shrinkage functions and evolutionary step sizes, thereby improving the effect of signal monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of an intelligent signal processing method provided by the present invention;

[0042] Figure 2 This is a schematic diagram of an intelligent signal processing system provided by the present invention.

[0043] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0045] In the description of the present invention, it should be understood that terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0046] Example 1, see Figure 1The present invention provides an intelligent signal processing method, which includes the following steps:

[0047] Step S1: signal acquisition: collecting historical electromagnetic signal data sets;

[0048] Step S2: signal preprocessing: performing denoising, pre-emphasis and window function segmentation on the signal;

[0049] Step S3: frequency domain analysis; converting the electromagnetic signal from the time domain to the frequency domain, extracting the frequency characteristics and removing the noise using a dynamic filter;

[0050] Step S4: local mean decomposition; decompose the signal by optimizing boundary conditions and stopping criteria to extract smooth trends and amplitude changes;

[0051] Step S5: mask signal processing; calculating the instantaneous frequency and using mask signal processing to filter signal components;

[0052] Step S6: Feature extraction: performing logarithmic transformation on the power spectrum of the signal, performing cepstrum analysis and enhancing the high-order features of the signal through a context-aware filter;

[0053] Step S7: Classification verification: classify the processed electromagnetic signal data based on a multi-class SVM and adjust the model through an optimization algorithm;

[0054] Step S8: Signal supervision: After real-time signal acquisition, signal processing and classification are performed based on the classification model, and warnings are issued for poor quality signals and relevant personnel are notified.

[0055] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the historical electromagnetic signal data set includes electromagnetic signal data and signal quality; the signal quality includes high quality, good, average and poor.

[0056] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the signal preprocessing is to remove noise and enhance the high-frequency characteristics of the electromagnetic signal; the received electromagnetic signal is pre-emphasized to amplify the high-frequency component and reduce the influence of the low-frequency component, which is expressed as: ; Use the window function to split the electromagnetic signal frame and reduce spectrum leakage; expressed as: ; ; ;in, It is the electromagnetic signal after pre-emphasis; is the original collected electromagnetic signal; is the dynamic pre-emphasis coefficient; is the basic pre-emphasis coefficient; t is the time; is the aggravation adjustment factor; is the previous sampling point value of the electromagnetic signal; n1 is the sampling point index; is the windowing adjustment factor.

[0057] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, the frequency domain analysis is to extract the frequency characteristics of the electromagnetic signal and remove the noise frequency; each frame of the electromagnetic signal is converted from the time domain to the frequency domain based on the fast Fourier transform, and the amplitude spectrum is calculated, which is expressed as: ; Use a triangular filter to map the linear frequency to the Mel frequency scale, expressed as: ; Perform dynamic filter adjustment, expressed as: ; ; The overall filter response is expressed as: ;in, is the amplitude spectrum of the frequency domain signal; N1 is the total number of frequencies; k is the frequency index; is the Fast Fourier Transform; is the mth Mel frequency eigenvalue; log is the logarithmic operation; is the frequency response of the triangular filter; is the filter weight after dynamic adjustment; are the initial filter weights; is the Sigmoid function, x is the function variable; is the input signal; is the signal-to-noise ratio; is the signal-to-noise ratio threshold; is the overall frequency response of the dynamic filter; N2 is the total number of filters; i is the filter index; is the dynamic adjustment weight of the filter at time t; is the frequency response of the i-th filter at the k-th frequency; is the noise power spectral density; is the maximum value of noise; is the filter adjustment coefficient.

[0058] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, local mean decomposition is performed. By optimizing boundary conditions, envelope estimation, and screening stop criteria, decomposition efficiency is improved and errors are reduced. The extreme local mean and amplitude are calculated, which is expressed as: ; ; The sliding average calculates the local mean and amplitude, which is expressed as: ; ; ; ; The stopping criterion value f is expressed as: ; The preliminary reconstructed signal is expressed as: ;in, It is the local mean of adjacent extreme points that reflects the smooth trend of the electromagnetic reflection signal; It is the local amplitude of adjacent extreme points, reflecting the amplitude change of electromagnetic signal; and are adjacent extreme value points in the electromagnetic signal; It is the result of removing the local mean of the electromagnetic signal; It is the normalized vibration component of the electromagnetic signal; is the local mean in the sliding window; is the local amplitude in the sliding window; is the input signal at time t; k is the length of the sliding window; is the total number of sampling points of the electromagnetic signal; n is the sampling point index of the electromagnetic signal; z is the error signal after the current decomposition; is the nth sampling point value of the signal residual; z1 is the mean value of the signal residual; is the threshold of the stopping criterion; is the signal after preliminary reconstruction; is the eigenmode function; is the decomposed residual signal, j is the total number of IMFs; p is the IMF index; is the amplitude adjustment factor.

[0059] By performing the above operations, this solution introduces dynamic pre-emphasis coefficients and window functions for signal preprocessing to address the problems of improper noise processing, insufficient distinction between noise and signal components, and serious deviation in reconstructed signals in general signal processing methods, which in turn lead to low accuracy in signal monitoring. Frequency domain analysis is performed based on dynamic filtering adjustment, and a dynamic filter is used to adjust the spectrum to remove noise and retain valid signals. By optimizing boundary conditions, the decomposition efficiency is improved and errors are reduced, thereby improving the accuracy of signal monitoring.

[0060] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the mask signal processing is to suppress modal aliasing by the mask signal and improve the decomposition accuracy; the instantaneous frequency is calculated using the Hilbert transform, which is expressed as: ; Mask signal Expressed as: ; Filter the decomposed components to remove noise and retain the effective electromagnetic signal; the component definition is expressed as: ; The final reconstructed electromagnetic signal Expressed as: ;in, is the instantaneous frequency at time t ; is the phase function of the electromagnetic signal; is the instantaneous amplitude of the component; is the effective modal component after screening; and are the component signals of the positive frequency and negative frequency parts respectively; n3 is the number of effective modal components screened out, and i3 is the corresponding index; is the filtered modal component.

[0061] Example 7, see Figure 1 This embodiment is based on the above embodiment. In step S6, feature extraction is to take the logarithm of the power spectrum, compress the dynamic range, and enhance small changes, which can be expressed as: ; Perform inverse Fourier transform on the power spectrum logarithm to obtain the cepstral coefficients, which are expressed as: ; Perform cepstral appreciation to enhance the high-order features in the cepstral coefficients to improve the electromagnetic signal classification and discrimination, which is expressed as: ; Introduce context-aware filters to adjust frequency characteristics for different applications, expressed as: ; ; The adjusted cepstral coefficient is expressed as: ; ;in, is the power spectrum of the signal; is the spectrum of the signal; are the cepstral coefficients; is the inverse Fourier transform; are the cepstral coefficients after appreciation; is the original value of the cepstral coefficient; n2 is the order of the cepstral coefficient; L is the length of the rise; is the frequency response of the context-aware filter; is the attenuation coefficient; d is the adjustment parameter; exp(·) is the exponential function; is the adjusted cepstral coefficient; is the basic attenuation coefficient; is the attenuation adjustment factor; f is the current frequency; is the center frequency; It is a random number between 0 and 1, independent each time; is the basic adjustment parameter; is the modulation amplitude control factor.

[0062] Example 8, see Figure 1 This embodiment is based on the above embodiment. In step S7, the classification verification is to classify the processed electromagnetic signal data based on a multi-class SVM, build a signal classification model, and optimize the signal classification model. Specifically, it includes:

[0063] Step S71: Based on the emphasis adjustment factor, adjustment parameter, filter adjustment coefficient, windowing adjustment factor, attenuation adjustment factor, modulation amplitude control factor, amplitude adjustment factor, kernel function and penalty parameter in the multi-class SVM, a parameter optimization space is established, and the individual position in the optimization population is initialized. The classification accuracy obtained based on the individual position is used as the individual fitness value; the individual position initialization is expressed as: ;in, and are the initialization positions of the o+1th individual and the oth individual respectively; p is the control parameter;

[0064] Step S72: Individual location update is expressed as:

[0065] ;

[0066] ;

[0067] ;

[0068] ;

[0069] in, The iteration position of the individual c-th dimension, T is the number of iterations, and Tmax is the maximum number of iterations; is the global optimal solution; is the correction parameter; is the position update factor; is the position shrinkage function; rand is a random number between 0 and 1; is the position of a random individual in the population; is the evolutionary step length; is the percentage difference from the global optimal solution; is a regulating factor; is the minimum weight; is the disturbance factor; is the control factor; is the convergence term; is the maximum development factor; is the variance of the current distribution of the population; is the maximum variance of the initial distribution of the population; is the gradient adjustment factor; is the fitness value gradient;

[0070] Step S73: A fitness threshold is set in advance. When an individual fitness value is higher than the fitness threshold, the optimization is complete, and a signal classification model is established based on the individual position. If the maximum number of iterations is reached, return to step S71; otherwise, continue with position iteration.

[0071] By performing the above operations, we can address the problems of general signal processing methods such as improper signal screening and inability to accurately remove noise components; ignoring high-order features in the signal, resulting in low signal classification accuracy and poor signal monitoring effect. This scheme enhances high-order features through cepstrum appreciation, improves the classification and discrimination of signals, especially the sensitivity to small changes; designs a context-aware filter to make feature extraction more targeted and flexible; and optimizes the signal classification model based on the design of position update factors, position shrinkage functions and evolutionary step sizes, thereby improving the effect of signal monitoring.

[0072] Example 9, see Figure 1 This embodiment is based on the above embodiment. In step S8, signal supervision is based on the established signal classification model, and electromagnetic signals are collected in real time. After processing in steps S2 to S6, classification processing is performed based on the signal classification model. Signals with general signal quality are marked, and an early warning is issued when the quality is reduced to poor; a warning is issued for signals directly classified as poor, and relevant personnel are notified to conduct inspection and maintenance.

[0073] Example 10, see Figure 2 , this embodiment is based on the above embodiment. The present invention provides an intelligent signal processing system, including a signal acquisition module, a signal preprocessing module, a frequency domain analysis module, a local mean decomposition module, a mask signal processing module, a feature extraction module, a classification verification module and a signal supervision module;

[0074] The signal acquisition module collects historical electromagnetic signal data sets and sends the data to the signal preprocessing module;

[0075] The signal preprocessing module performs denoising, pre-emphasis and window function segmentation on the signal; and sends the data to the frequency domain analysis module;

[0076] The frequency domain analysis module converts the electromagnetic signal from the time domain to the frequency domain, extracts the frequency characteristics and removes the noise using a dynamic filter; and sends the data to the local mean decomposition module;

[0077] The local mean decomposition module decomposes the signal by optimizing the boundary conditions and the stopping criteria, extracts the smoothing trend and the amplitude change, and sends the data to the mask signal processing module;

[0078] The mask signal processing module calculates the instantaneous frequency and uses mask signal processing to filter the signal components; and sends the data to the feature extraction module;

[0079] The feature extraction module performs logarithmic transformation on the power spectrum of the signal, performs cepstrum analysis and enhances the high-order features of the signal through a context-aware filter; and sends the data to the classification verification module;

[0080] The classification verification module classifies the processed electromagnetic signal data based on a multi-class SVM and adjusts the model through an optimization algorithm; and sends the data to the signal supervision module;

[0081] After real-time signal acquisition, the signal monitoring module processes and classifies the signals based on the classification model, issues early warnings for signals of poor quality, and notifies relevant personnel.

[0082] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0083] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0084] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. An intelligent signal processing method, characterized in that: The method comprises the following steps: Step S1: signal acquisition; Step S2: signal preprocessing; Step S3: frequency domain analysis; Step S4: local mean decomposition; Step S5: mask signal processing; Step S6: feature extraction; Step S7: classification verification; Step S8: signal supervision; In step S2, the signal preprocessing is to remove noise and enhance the high-frequency characteristics of the electromagnetic signal; the received electromagnetic signal is pre-emphasized, which is expressed as: ; Use the window function to split the electromagnetic signal frame; expressed as: ; ; ;in, It is the electromagnetic signal after pre-emphasis; is the original collected electromagnetic signal; is the dynamic pre-emphasis coefficient, t is the time; is the basic pre-emphasis coefficient; is the aggravation adjustment factor; is the previous sampling point value of the electromagnetic signal; n1 is the sampling point index; is the windowing adjustment factor; In step S4, the local mean decomposition is the extreme local mean and amplitude calculation, which is expressed as: ; ; The sliding average calculates the local mean and amplitude, which is expressed as: ; ; ; ; The stopping criterion value f is expressed as: ; The preliminary reconstructed signal is expressed as: ;in, It is the local mean of adjacent extreme points that reflects the smooth trend of the electromagnetic reflection signal; It is the local amplitude of adjacent extreme points, reflecting the amplitude change of electromagnetic signal; and are adjacent extreme value points in the electromagnetic signal; It is the result of removing the local mean of the electromagnetic signal; It is the normalized vibration component of the electromagnetic signal; is the local mean in the sliding window; is the local amplitude in the sliding window; is the input signal at time t; k is the length of the sliding window; is the total number of sampling points of the electromagnetic signal; n is the sampling point index of the electromagnetic signal; z is the error signal after the current decomposition; is the nth sampling point value of the signal residual; z1 is the mean value of the signal residual; is the threshold of the stopping criterion; is the signal after preliminary reconstruction; is the eigenmode function; is the decomposed residual signal, j is the total number of IMFs; p is the IMF index; is the amplitude adjustment factor; In step S6, the feature extraction is to take the logarithm of the power spectrum, which is expressed as: ; Perform inverse Fourier transform on the power spectrum logarithm to obtain the cepstral coefficients, which are expressed as: ; Perform cepstral appreciation to enhance the high-order features in the cepstral coefficients, expressed as: ; Introduce context-aware filter, expressed as: ; ; The adjusted cepstral coefficient is expressed as: ; ;in, is the power spectrum of the signal; is the spectrum of the signal; are the cepstral coefficients; is the inverse Fourier transform; are the cepstral coefficients after appreciation; is the original value of the cepstral coefficient; n2 is the order of the cepstral coefficient; L is the length of the rise; is the frequency response of the context-aware filter; is the attenuation coefficient; d is the adjustment parameter; exp(·) is the exponential function; is the adjusted cepstral coefficient; is the basic attenuation coefficient; is the attenuation adjustment factor; f is the current frequency; is the center frequency; It is a random number between 0 and 1, independent each time; is the basic adjustment parameter; is the modulation amplitude control factor; In step S7, the classification verification is to classify the processed electromagnetic signal data based on a multi-class SVM, construct a signal classification model, and optimize the signal classification model; including: Step S71: Based on the emphasis adjustment factor, adjustment parameter, filter adjustment coefficient, windowing adjustment factor, attenuation adjustment factor, modulation amplitude control factor, amplitude adjustment factor, kernel function and penalty parameter in the multi-class SVM, a parameter optimization space is established, and the individual position in the optimization population is initialized. The classification accuracy obtained based on the individual position is used as the individual fitness value; the individual position initialization is expressed as: ;in, and are the initialization positions of the o+1th individual and the oth individual respectively; p is the control parameter; Step S72: Individual location update is expressed as: ； ； ； ； in, The iteration position of the individual c-th dimension, T is the number of iterations, and Tmax is the maximum number of iterations; is the global optimal solution; is the correction parameter; is the position update factor; is the position shrinkage function; rand is a random number between 0 and 1; is the position of a random individual in the population; is the evolutionary step length; is the percentage difference from the global optimal solution; is a regulating factor; is the minimum weight; is the disturbance factor; is the control factor; is the convergence term; is the maximum development factor; is the variance of the current distribution of the population; is the maximum variance of the initial distribution of the population; is the gradient adjustment factor; is the fitness value gradient; Step S73: A fitness threshold is set in advance. When an individual fitness value is higher than the fitness threshold, the optimization is complete, and a signal classification model is established based on the individual position. If the maximum number of iterations is reached, return to step S71; otherwise, continue with position iteration.

2. The intelligent signal processing method according to claim 1, characterized in that: In step S3, the frequency domain analysis is to convert each frame of electromagnetic signal from the time domain to the frequency domain based on the fast Fourier transform, and calculate the amplitude spectrum, which is expressed as: ; Use a triangular filter to map the linear frequency to the Mel frequency scale, expressed as: ; Perform dynamic filter adjustment, expressed as: ; ; The overall filter response is expressed as: ;in, is the amplitude spectrum of the frequency domain signal; N1 is the total number of frequencies; k is the frequency index; is the Fast Fourier Transform; is the mth Mel frequency eigenvalue; log is the logarithmic operation; is the frequency response of the triangular filter; is the filter weight after dynamic adjustment; are the initial filter weights; is the Sigmoid function, x is the function variable; is the input signal; is the signal-to-noise ratio; is the signal-to-noise ratio threshold; is the overall frequency response of the dynamic filter; N2 is the total number of filters; i is the filter index; is the dynamic adjustment weight of the filter at time t; is the frequency response of the i-th filter at the k-th frequency; is the noise power spectral density; is the maximum value of noise; is the filter adjustment coefficient.

3. The intelligent signal processing method according to claim 2, characterized in that: In step S5, the mask signal processing is to calculate the instantaneous frequency using Hilbert transform, which is expressed as: ; Mask signal Expressed as: ; Filter the decomposed components to remove noise and retain the effective electromagnetic signal; the component definition is expressed as: ; The final reconstructed electromagnetic signal Expressed as: ;in, is the instantaneous frequency at time t ; is the phase function of the electromagnetic signal; is the instantaneous amplitude of the component; is the effective modal component after screening; and are the component signals of the positive frequency and negative frequency parts respectively; n3 is the number of effective modal components screened out, and i3 is the corresponding index; is the filtered modal component.

4. The intelligent signal processing method according to claim 3, characterized in that: In step S1 , the signal acquisition is to acquire a historical electromagnetic signal data set; the historical electromagnetic signal data set includes electromagnetic signal data and signal quality; the signal quality includes excellent, good, average and poor.

5. The intelligent signal processing method according to claim 4, characterized in that: In step S8, the signal monitoring is based on the established signal classification model, and electromagnetic signals are collected in real time. After being processed in steps S2 to S6, they are classified based on the signal classification model, and signals with average signal quality are marked. When the quality deteriorates to poor, an early warning is issued; Signals directly classified as poor will be warned and relevant personnel will be notified to conduct inspection and maintenance.

6. An intelligent signal processing system for implementing an intelligent signal processing method according to any one of claims 1 to 5, characterized in that: It includes signal acquisition module, signal preprocessing module, frequency domain analysis module, local mean decomposition module, mask signal processing module, feature extraction module, classification verification module and signal supervision module; The signal acquisition module collects historical electromagnetic signal data sets and sends the data to the signal preprocessing module; The signal preprocessing module performs denoising, pre-emphasis and window function segmentation on the signal; and sends the data to the frequency domain analysis module; The frequency domain analysis module converts the electromagnetic signal from the time domain to the frequency domain, extracts the frequency characteristics and removes the noise using a dynamic filter; and sends the data to the local mean decomposition module; The local mean decomposition module decomposes the signal by optimizing the boundary conditions and the stopping criteria, extracts the smoothing trend and the amplitude change, and sends the data to the mask signal processing module; The mask signal processing module calculates the instantaneous frequency and uses mask signal processing to filter the signal components; and sends the data to the feature extraction module; The feature extraction module performs logarithmic transformation on the power spectrum of the signal, performs cepstrum analysis and enhances the high-order features of the signal through a context-aware filter; and sends the data to the classification verification module; The classification verification module classifies the processed electromagnetic signal data based on a multi-class SVM and adjusts the model through an optimization algorithm; and sends the data to the signal supervision module; After real-time signal acquisition, the signal monitoring module processes and classifies the signals based on the classification model, issues early warnings for signals of poor quality, and notifies relevant personnel.

Images