A method for adjusting the sampling frequency of myoelectric signals

Abstract

The application relates to the technical field of data processing, and discloses a method for adjusting the sampling frequency of myoelectric signals. The method collects the electric signal data generated by muscles in real time through an installed electrode detection device, matrix-constructs an electric signal data matrix, simultaneously processes the collected electric signal data matrix through a data processing mode, simultaneously analyzes the processed electric signal data matrix through a data analysis method, extracts the features of the analyzed electric signal data through a feature extraction algorithm, collects the electric signal data after feature extraction to construct an SVM model, simultaneously trains the constructed SVM model through a training optimization method, obtains a final SVM model, classifies the electric signal data after feature extraction based on the final SVM model, and finally constructs an electric signal data grading system, thereby improving the accuracy of myoelectric signal sampling.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method for adjusting the sampling frequency of electromyography (EMG) signals. Background Technology

[0002] As surface electromyography (SEMG) technology becomes increasingly prevalent in rehabilitation medicine, sports training, and human-computer interaction, there is a need for a practical, portable, high-quality, low-cost, and easy-to-use SEMG signal acquisition and processing device. However, due to the variability of SEMG signals, signal acquisition is easily affected by various factors.

[0003] Existing technology, such as the invention patent application with publication number CN114668563B, discloses a multi-level adjustment method for the sampling frequency of electromyography (EMG) signals. This invention acquires a first EMG signal. When it is determined that the target action corresponding to the first EMG signal is a continuous action, the EMG sampling frequency of the bionic hand is down-adjusted according to a first value. It acquires a second EMG signal. When it is determined that the second EMG signal is the target EMG signal, the cumulative number of consecutive acquisitions of the target EMG signal is acquired. The second EMG signal is the EMG signal acquired after the first EMG signal, and the target EMG signal is one of several high-frequency EMG signals. An up-adjustment value is determined based on the cumulative number of consecutive acquisitions, and the EMG sampling frequency is up-adjusted according to the up-adjustment value.

[0004] As can be seen from the above solutions, current sampling of electromyographic (EMG) signals is often done by identifying and then acquiring the data by frequency modulation, which ignores the variability of EMG signals and limits the amount of data collected. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method for adjusting the sampling frequency of electromyography (EMG) signals, which has the advantages of accuracy, real-time performance, and high efficiency, and solves the problem that signal acquisition is easily affected by various factors due to the variability of surface EMG signals.

[0007] (II) Technical Solution

[0008] To address the aforementioned technical problem that the acquisition of surface electromyography (EMG) signals is easily affected by various factors due to the variability of these signals, this invention provides the following technical solution:

[0009] This embodiment discloses a method for adjusting the sampling frequency of electromyography signals, specifically including the following steps:

[0010] S1. Install detection electrode devices on the surface of the muscle and collect electrical signal data generated by the muscle in real time through the installed electrode detection devices; at the same time, obtain the collected electrical signal data matrix through matrix construction method;

[0011] S2. The collected electrical signal data matrix is ​​processed using data processing methods to obtain the processed electrical signal data matrix.

[0012] S21. Remove noise from the acquired electrical signal data matrix using data denoising methods to obtain the denoised electrical signal data matrix;

[0013] S22. The noise-reduced electrical signal data is filtered by data filtering to obtain the filtered electrical signal data matrix.

[0014] S23. Set the filtered electrical signal data matrix as the processed electrical signal data matrix;

[0015] S3. Analyze the processed electrical signal data matrix using data analysis methods to obtain the analyzed electrical signal data; simultaneously, extract features from the analyzed electrical signal data using a feature extraction algorithm and output the feature-extracted electrical signal data.

[0016] S4. Summarize the feature-extracted electrical signal data to construct an SVM model. Simultaneously, train the constructed SVM model using training optimization methods to obtain the final SVM model. Based on the final SVM model, classify the feature-extracted electrical signal data to obtain classified electrical signal data.

[0017] S5. Construct an electrical signal data hierarchy system based on the characteristics of electrical signals, and perform frequency modulation management on the classified electrical signal data based on the constructed electrical signal data hierarchy system.

[0018] Preferably, the step of installing detection electrode devices on the muscle surface and acquiring electrical signal data generated by the muscle in real time through the installed electrode detection devices, while obtaining the acquired electrical signal data matrix through a matrix construction method, includes the following steps:

[0019] S11. Each acquisition point in the installed electrode detection device includes two sets of electrodes and a differential amplifier; the electrical signals acquired by the two sets of electrodes are set as a set of real-time acquisition of electrical signal data generated by the muscle; after the electrical signal data acquisition is completed, the acquired electrical signal data will be amplified by the differential amplifier;

[0020] Set the collection time points of the electrode detection device , Indicates the set number Each data collection time point This indicates the total number of data collection time points set.

[0021] S12. Summarize the electrical signal data collected at each set acquisition time point to obtain a first data matrix; and use the first data as the electrical signal data matrix, as shown below:

[0022] ;

[0023] in, Indicating the first in the electrode detection equipment The collection point is at the [number]th [location]. Data collected at various time points.

[0024] Preferably, the step of removing noise from the acquired electrical signal data matrix through data denoising to obtain the denoised electrical signal data matrix includes the following steps:

[0025] S211. Transform the electrical signal data using a wavelet transform function;

[0026] The wavelet transform function is shown below:

[0027] ;

[0028] in, As a scale factor, It is the displacement factor. It is a square-integrable function. For the mother wavelet, It is a wavelet transform function;

[0029] S212. After the transformation is completed, a threshold is set to perform noise reduction processing on the electrical signal data in the electrical signal data matrix.

[0030] The noise expression for the electrical signal data in the electrical signal data matrix is ​​shown below:

[0031] ;

[0032] in, The electrical signal data in the electrical signal data matrix. Let the standard deviation of the noise be . For noise;

[0033] The wavelet-transformed electrical signal data is divided by a threshold function, and the electrical signal data that is less than the threshold is adjusted to 0.

[0034] Threshold function:

[0035] ;

[0036] Where λ is the set threshold, Let j represent the amplitude of the electrical signal data, k represent the horizontal axis of the amplitude, and sign represent the sign function. This is the noise-reduced electrical signal data.

[0037] Preferably, the step of filtering the denoised electrical signal data to obtain the filtered electrical signal data matrix includes the following steps:

[0038] The signal-to-noise ratio is set as the noise reduction evaluation standard, and the noise-reduced electrical signal data is filtered based on the set noise reduction evaluation standard.

[0039] The signal-to-noise ratio formula is as follows:

[0040] ;

[0041] in, Indicates the signal-to-noise ratio. This represents the electrical signal data after noise reduction;

[0042] The higher the signal-to-noise ratio, the stronger the noise suppression effect.

[0043] Set a signal-to-noise ratio (SNR) threshold. If the SNR calculated from the denoised electrical signal data is less than the set SNR threshold, the denoising result is unqualified. The denoised electrical signal data is then deleted and filtered. If the SNR calculated from the denoised electrical signal data is greater than or equal to the set SNR threshold, the denoising result is qualified.

[0044] The filtered electrical signal data are aggregated to form a filtered electrical signal data matrix.

[0045] Preferably, the step of analyzing the processed electrical signal data matrix using data analysis methods to obtain analyzed electrical signal data, and simultaneously extracting features from the analyzed electrical signal data using a feature extraction algorithm to output feature-extracted electrical signal data, includes the following steps:

[0046] S31. The processed electrical signal data matrix is ​​analyzed using data analysis methods to obtain the analyzed electrical signal data.

[0047] S32. After determining the amplitude range of the electrical signal data in the processed electrical signal data matrix, feature extraction is performed on the analyzed electrical signal data using a feature extraction algorithm.

[0048] Preferably, the step of analyzing the processed electrical signal data matrix using data analysis methods to obtain the analyzed electrical signal data includes the following steps:

[0049] Amplitude keying modulation (ASK) is used to perform modulation analysis on the processed electrical signal data matrix to determine the amplitude range of the electrical signal data in the processed electrical signal data matrix.

[0050] The time-domain expression for amplitude shift keying (APS) modulation is:

[0051] ;

[0052] Among them, electrical signals Pulse width is , For the time domain of amplitude shift keying modulation, Let η be the value of the ηth electrical signal. It is represented as the carrier frequency of the k-th electrical signal;

[0053] The electrical signal data after amplitude keying modulation analysis is obtained by summing the time domain data of amplitude keying modulation.

[0054] ;

[0055] Furthermore, the amplitude range of the electrical signal data in the processed electrical signal data matrix is ​​determined by summarizing the electrical signal data after amplitude keying modulation analysis.

[0056] Preferably, after determining the amplitude range of the electrical signal data in the processed electrical signal data matrix, the step of simultaneously extracting features from the analyzed electrical signal data using a feature extraction algorithm includes the following steps:

[0057] S321. Divide the electrical signal data in the processed electrical signal data matrix into K data blocks of uniform size, and use the divided data blocks as the input of the convolutional neural network.

[0058] The convolutional neural network includes: convolutional layers, pooling layers, and fully connected layers;

[0059] S322. Based on the divided data blocks, set the convolution kernel size and move it on the input image data blocks according to the set stride. During the movement, perform convolution calculation on the data in the corresponding region data blocks to achieve feature extraction for each data block.

[0060] The formula for calculating convolution is as follows:

[0061] ;

[0062] in, This represents the input data block. represents the weights of the corresponding convolution kernel, b represents the bias value, and F represents the output feature;

[0063] Furthermore, the convolutional and pooling layers are stacked continuously, and the output of the upper layer becomes the input of the lower layer. After all the convolution and pooling operations are completed, the output image data block features are passed into the fully connected layer.

[0064] The output image data block features are expanded and combined using a fully connected layer to obtain a single feature data, which is then saved.

[0065] Preferably, the process of summarizing the extracted electrical signal data to construct an SVM model, training the constructed SVM model using a training optimization method to obtain a final SVM model, and then classifying the extracted electrical signal data based on the final SVM model to obtain classified electrical signal data includes the following steps:

[0066] S41. Construct the initial SVM model; set the kernel function used in the initial SVM model to be the Gaussian kernel function; as follows.

[0067] The mathematical expression for the Support Vector Machine algorithm is shown below:

[0068] ;

[0069] Where w represents the weight and b represents the bias value. Indicates sample The classification results In the formula, Represents the Gaussian kernel function; This represents the data in the SVM classification space; This represents the core of the Gaussian kernel function; express and The Euclidean distance between them; Indicates the scope of the Gaussian kernel function;

[0070] S42. Set the training data ratio as follows: According to the proportion of the training data The extracted electrical signal data is divided into training and testing data matrices; the maximum number of iterations is set to... and the training error threshold is ;

[0071] The training data matrix is ​​input into the initial SVM model to train the initial SVM model. When the number of training iterations is greater than or equal to... When or the training error is less than When the initial SVM model is stopped, the trained SVM model is obtained.

[0072] S43. Set a test accuracy threshold, input the test data matrix into the trained SVM model for classification, and obtain the classification result; compare the classification result with the classification label corresponding to the pre-set test data matrix to obtain the classification accuracy; when the classification accuracy is greater than or equal to the test accuracy threshold, use the trained SVM model as the final SVM model; otherwise, repeat S41 and S42 until the classification accuracy is greater than or equal to the test accuracy threshold.

[0073] Preferably, the step of constructing an electrical signal data hierarchy based on the characteristics of electrical signals, and performing frequency modulation management on the classified electrical signal data based on the constructed electrical signal data hierarchy, includes the following steps:

[0074] Based on the amplitude range of electromuscular signals, an electrical signal data grading system is established, with each 0.1mV representing one grade.

[0075] Set a frequency modulation management threshold for electrical signals. When the classified electrical signal data is lower than the set evaluation threshold, increase the sampling frequency of the electrical signal data. When the classified electrical signal data is higher than the set evaluation threshold, decrease the sampling frequency of the electrical signal data.

[0076] This embodiment also discloses a system for adjusting the sampling frequency of electromyography (EMG) signals, which is used to implement a method for adjusting the sampling frequency of EMG signals. The system includes: a data acquisition device, a data processing module, a data analysis module, a feature extraction module, a classification module, and a frequency modulation management module.

[0077] The data acquisition device includes a detection electrode device and a differential amplifier, used to acquire and amplify electrical signal data generated by muscles in real time;

[0078] The data processing module is used to process the real-time acquired electrical signal data to obtain processed electrical signal data.

[0079] The data analysis module is used to analyze the processed electrical signal data;

[0080] The feature extraction module is used to extract features from the analyzed electrical signal data;

[0081] The classification module is used to construct a classification model based on the feature-extracted electrical signal data, and to classify the electrical signal data based on the constructed classification module.

[0082] The frequency modulation management module is used to construct an electrical signal data hierarchy system, classify the classified electrical signal data into different levels, and perform frequency modulation management based on the classification results.

[0083] (III) Beneficial Effects

[0084] Compared with the prior art, the present invention provides a method for adjusting the sampling frequency of electromyography signals, which has the following beneficial effects:

[0085] 1. This invention collects electrical signal data generated by muscles in real time through an installed electrode detection device, constructs an electrical signal data matrix, processes the collected electrical signal data matrix through data processing methods, analyzes the processed electrical signal data matrix through data analysis methods, and extracts features from the analyzed electrical signal data through a feature extraction algorithm. After extraction, the feature-extracted electrical signal data is summarized to construct an SVM model, and the constructed SVM model is trained through a training optimization method to obtain the final SVM model. Based on the final SVM model, the feature-extracted electrical signal data is classified, and finally, an electrical signal data hierarchy system is constructed. Based on the constructed electrical signal data hierarchy system, the classified electrical signal data is frequency-modulated, thereby improving the accuracy of electromyography signal sampling.

[0086] 2. This invention acquires electromyographic signals by using two sets of electrodes and a differential amplifier at each acquisition point in the installed electrode detection device. At the same time, it determines the electrical signal data matrix by setting the acquisition time point and constructing the data matrix, thereby quantifying the acquisition of electrical signal data.

[0087] 3. This invention removes noise from the acquired electrical signal data matrix by using wavelet function denoising to obtain a denoised electrical signal data matrix. At the same time, it filters the electrical signal data in the denoised electrical signal data matrix by using the signal-to-noise ratio as the denoising evaluation standard, thereby improving the reliability of electrical signal data processing.

[0088] 4. This invention analyzes the processed electrical signal data matrix by using amplitude keying modulation and determines the amplitude range of the electrical signal data in the processed electrical signal data matrix. After determining the amplitude range, feature extraction is performed on the electrical signal data in the processed electrical signal data matrix, thereby improving the accuracy of electrical signal feature extraction.

[0089] 5. This invention constructs an SVM model by summarizing the electrical signal data after feature extraction, and trains the constructed SVM model through a training optimization method to obtain the final SVM model. Based on the final SVM model, the electrical signal data after feature extraction is classified, thereby improving the intelligence of electrical signal data classification. Attached Figure Description

[0090] Figure 1 This is a schematic diagram of the electromyography signal sampling frequency adjustment method of the present invention. Detailed Implementation

[0091] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0092] Example 1

[0093] Please see Figure 1 This embodiment discloses a method for adjusting the sampling frequency of electromyography signals, specifically including the following steps:

[0094] S1. Install detection electrode devices on the surface of the muscle and collect electrical signal data generated by the muscle in real time through the installed electrode detection devices; at the same time, obtain the collected electrical signal data matrix through matrix construction method;

[0095] S2. The collected electrical signal data matrix is ​​processed using data processing methods to obtain the processed electrical signal data matrix.

[0096] S21. Remove noise from the acquired electrical signal data matrix using data denoising methods to obtain the denoised electrical signal data matrix;

[0097] S22. The noise-reduced electrical signal data is filtered by data filtering to obtain the filtered electrical signal data matrix.

[0098] S23. Set the filtered electrical signal data matrix as the processed electrical signal data matrix;

[0099] S3. Analyze the processed electrical signal data matrix using data analysis methods to obtain the analyzed electrical signal data; simultaneously, extract features from the analyzed electrical signal data using a feature extraction algorithm and output the feature-extracted electrical signal data.

[0100] S4. Summarize the feature-extracted electrical signal data to construct an SVM model. Simultaneously, train the constructed SVM model using training optimization methods to obtain the final SVM model. Based on the final SVM model, classify the feature-extracted electrical signal data to obtain classified electrical signal data.

[0101] S5. Construct an electrical signal data hierarchy system based on the characteristics of electrical signals, and perform frequency modulation management on the classified electrical signal data based on the constructed electrical signal data hierarchy system;

[0102] Further, please refer to Figure 1The process of installing detection electrode devices on the muscle surface and acquiring electrical signal data generated by the muscle in real time through the installed electrode detection devices, while obtaining the acquired electrical signal data matrix through a matrix construction method, includes the following steps:

[0103] S11. Each acquisition point in the installed electrode detection device includes two sets of electrodes and a differential amplifier; the electrical signals acquired by the two sets of electrodes are set as a set of real-time acquisition of electrical signal data generated by the muscle; after the electrical signal data acquisition is completed, the acquired electrical signal data will be amplified by the differential amplifier;

[0104] Furthermore, a set of data acquisition time points for the electrode detection device is defined. , Indicates the set number Each data collection time point This indicates the total number of data collection time points set.

[0105] Furthermore, the expression for the acquired electrical signal is as follows:

[0106] ;

[0107] in, This represents a set of electrical signal data generated by the muscles collected at time t. This represents the electrical signal acquired by the electrode at time t. This indicates the time it takes for an electrical signal to travel from one electrode to another.

[0108] S12. Summarize the electrical signal data collected at each set acquisition time point to obtain a first data matrix; and use the first data as the electrical signal data matrix, as shown below:

[0109] ;

[0110] in, Indicating the first in the electrode detection equipment The collection point is at the [number]th [location]. Data collected at each time point;

[0111] Further, please refer to Figure 1 The process of removing noise from the acquired electrical signal data matrix using data denoising methods to obtain the denoised electrical signal data matrix includes the following steps:

[0112] S211. Transform the electrical signal data using a wavelet transform function;

[0113] The wavelet transform function is shown below:

[0114] ;

[0115] in, As a scale factor, It is the displacement factor. It is a square-integrable function. For the mother wavelet, It is a wavelet transform function;

[0116] S212. After the transformation is completed, a threshold is set to perform noise reduction processing on the electrical signal data in the electrical signal data matrix.

[0117] The noise expression for the electrical signal data in the electrical signal data matrix is ​​shown below:

[0118] ;

[0119] in, The electrical signal data in the electrical signal data matrix. Let the standard deviation of the noise be . For noise;

[0120] The wavelet-transformed electrical signal data is divided by a threshold function, and the electrical signal data that is less than the threshold is adjusted to 0.

[0121] Threshold function:

[0122] ;

[0123] Where λ is the set threshold, Let j represent the amplitude of the electrical signal data, k represent the horizontal axis of the amplitude, and sign represent the sign function. The data is the noise-reduced electrical signal data;

[0124] Further, please refer to Figure 1 The process of filtering the denoised electrical signal data to obtain the filtered electrical signal data matrix involves the following steps:

[0125] The signal-to-noise ratio is set as the noise reduction evaluation standard, and the noise-reduced electrical signal data is filtered based on the set noise reduction evaluation standard.

[0126] The signal-to-noise ratio formula is as follows:

[0127] ;

[0128] in, Indicates the signal-to-noise ratio. This represents the electrical signal data after noise reduction;

[0129] The higher the signal-to-noise ratio, the stronger the noise suppression effect.

[0130] Set a signal-to-noise ratio (SNR) threshold. If the SNR calculated from the denoised electrical signal data is less than the set SNR threshold, the denoising result is unqualified. The denoised electrical signal data is then deleted and filtered. If the SNR calculated from the denoised electrical signal data is greater than or equal to the set SNR threshold, the denoising result is qualified.

[0131] The filtered electrical signal data are summarized to form a filtered electrical signal data matrix;

[0132] Further, please refer to Figure 1 The processed electrical signal data matrix is ​​analyzed using data analysis methods to obtain the analyzed electrical signal data. Simultaneously, feature extraction algorithms are used to extract features from the analyzed electrical signal data, outputting the feature-extracted electrical signal data. The process includes the following steps:

[0133] S31. The processed electrical signal data matrix is ​​analyzed using data analysis methods to obtain the analyzed electrical signal data.

[0134] Amplitude keying modulation (ASK) is used to perform modulation analysis on the processed electrical signal data matrix to determine the amplitude range of the electrical signal data in the processed electrical signal data matrix.

[0135] The time-domain expression for amplitude shift keying (APS) modulation is:

[0136] ;

[0137] Among them, electrical signals Pulse width is , For the time domain of amplitude shift keying modulation, Let η be the value of the ηth electrical signal. It is represented as the carrier frequency of the k-th electrical signal;

[0138] The electrical signal data after amplitude keying modulation analysis is obtained by summing the time domain data of amplitude keying modulation.

[0139] ;

[0140] Furthermore, the amplitude range of the electrical signal data in the processed electrical signal data matrix is ​​determined by summarizing the electrical signal data after amplitude keying modulation analysis.

[0141] S32. After determining the amplitude range of the electrical signal data in the processed electrical signal data matrix, feature extraction is performed on the analyzed electrical signal data using a feature extraction algorithm.

[0142] S321. Divide the electrical signal data in the processed electrical signal data matrix into K data blocks of uniform size, and use the divided data blocks as the input of the convolutional neural network.

[0143] The convolutional neural network includes: convolutional layers, pooling layers, and fully connected layers;

[0144] S322. Based on the divided data blocks, set the convolution kernel size and move it on the input image data blocks according to the set stride. During the movement, perform convolution calculation on the data in the corresponding region data blocks to achieve feature extraction for each data block.

[0145] The formula for calculating convolution is as follows:

[0146] ;

[0147] in, This represents the input data block. represents the weights of the corresponding convolution kernel, b represents the bias value, and F represents the output feature;

[0148] Furthermore, the convolutional and pooling layers are stacked continuously, and the output of the upper layer becomes the input of the lower layer. After all the convolution and pooling operations are completed, the output image data block features are passed into the fully connected layer.

[0149] The output image data block features are expanded and combined using a fully connected layer to obtain a single feature data, which is then saved.

[0150] Further, please refer to Figure 1 The extracted electrical signal data is aggregated to construct an SVM model. Simultaneously, the constructed SVM model is trained using training optimization methods to obtain the final SVM model. Based on the final SVM model, the extracted electrical signal data is then classified to obtain the classified electrical signal data. The steps include:

[0151] S41. Construct the initial SVM model; set the kernel function used in the initial SVM model to be the Gaussian kernel function; as follows.

[0152] The mathematical expression for the Support Vector Machine algorithm is shown below:

[0153] ;

[0154] Where w represents the weight and b represents the bias value. Indicates sample The classification results In the formula, Represents the Gaussian kernel function; This represents the data in the SVM classification space; This represents the core of the Gaussian kernel function; express and The Euclidean distance between them; Indicates the scope of the Gaussian kernel function;

[0155] S42. Set the training data ratio as follows: According to the proportion of the training data The extracted electrical signal data is divided into training and testing data matrices; the maximum number of iterations is set to... and the training error threshold is ;

[0156] The training data matrix is ​​input into the initial SVM model to train the initial SVM model. When the number of training iterations is greater than or equal to... When or the training error is less than When the initial SVM model is stopped, the trained SVM model is obtained.

[0157] S43. Set a test accuracy threshold, input the test data matrix into the trained SVM model for classification, and obtain the classification result; compare the classification result with the classification label corresponding to the pre-set test data matrix to obtain the classification accuracy; when the classification accuracy is greater than or equal to the test accuracy threshold, use the trained SVM model as the final SVM model; otherwise, repeat S41 and S42 until the classification accuracy is greater than or equal to the test accuracy threshold.

[0158] Further, please refer to Figure 1 The process of constructing an electrical signal data hierarchy based on the characteristics of electrical signals, and then performing frequency modulation management on the classified electrical signal data based on the constructed electrical signal data hierarchy, includes the following steps:

[0159] Based on the amplitude range of electromuscular signals, an electrical signal data grading system is established, with each 0.1mV representing one grade.

[0160] Set a frequency modulation management threshold for electrical signals. When the classified electrical signal data is lower than the set evaluation threshold, increase the sampling frequency of the electrical signal data. When the classified electrical signal data is higher than the set evaluation threshold, decrease the sampling frequency of the electrical signal data.

[0161] Example 2

[0162] In one specific embodiment, the system for adjusting the sampling frequency of electromyography (EMG) signals is used to implement a method for adjusting the sampling frequency of EMG signals. The system includes: a data acquisition device, a data processing module, a data analysis module, a feature extraction module, a classification module, and a frequency modulation management module.

[0163] The data acquisition device includes a detection electrode device and a differential amplifier, used to acquire and amplify electrical signal data generated by muscles in real time;

[0164] The data processing module is used to process the real-time acquired electrical signal data to obtain processed electrical signal data.

[0165] The data analysis module is used to analyze the processed electrical signal data;

[0166] The feature extraction module is used to extract features from the analyzed electrical signal data;

[0167] The classification module is used to construct a classification model based on the feature-extracted electrical signal data, and to classify the electrical signal data based on the constructed classification module.

[0168] The frequency modulation management module is used to construct an electrical signal data hierarchy system, classify the classified electrical signal data into different levels, and perform frequency modulation management based on the classification results.

[0169] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for adjusting the sampling frequency of electromyographic signals, characterized in that, Includes the following steps: S1. Install detection electrode devices on the surface of the muscle and collect electrical signal data generated by the muscle in real time through the installed electrode detection devices; Simultaneously, the collected electrical signal data matrix is ​​obtained through matrix construction methods; S2. The collected electrical signal data matrix is ​​processed using data processing methods to obtain the processed electrical signal data matrix. S21. Remove noise from the acquired electrical signal data matrix using data denoising methods to obtain the denoised electrical signal data matrix; S22. The noise-reduced electrical signal data is filtered by data filtering to obtain the filtered electrical signal data matrix. S23. Set the filtered electrical signal data matrix as the processed electrical signal data matrix; S3. Analyze the processed electrical signal data matrix using data analysis methods to obtain the analyzed electrical signal data; simultaneously, extract features from the analyzed electrical signal data using a feature extraction algorithm and output the feature-extracted electrical signal data. S31. The processed electrical signal data matrix is ​​analyzed using data analysis methods to obtain the analyzed electrical signal data. Amplitude keying modulation (ASK) is used to perform modulation analysis on the processed electrical signal data matrix to determine the amplitude range of the electrical signal data in the processed electrical signal data matrix. The time-domain expression for amplitude shift keying (APS) modulation is: ； Among them, electrical signals Pulse width is , For the time domain of amplitude shift keying modulation, Let η be the value of the ηth electrical signal. It is represented as the carrier frequency of the k-th electrical signal; The electrical signal data after amplitude keying modulation analysis is obtained by summing the time domain data of amplitude keying modulation. ； Summarize the electrical signal data after amplitude keying modulation analysis to determine the amplitude range of the electrical signal data in the processed electrical signal data matrix; S32. After determining the amplitude range of the electrical signal data in the processed electrical signal data matrix, feature extraction is performed on the analyzed electrical signal data using a feature extraction algorithm. S321. Divide the electrical signal data in the processed electrical signal data matrix into K data blocks of uniform size, and use the divided data blocks as the input of the convolutional neural network. The convolutional neural network includes: convolutional layers, pooling layers, and fully connected layers; S322. Based on the divided data blocks, set the convolution kernel size and move it on the input image data blocks according to the set stride. During the movement, perform convolution calculation on the data in the corresponding region data blocks to achieve feature extraction for each data block. The formula for calculating convolution is as follows: ； in, This represents the input data block. represents the weights of the corresponding convolution kernel, b represents the bias value, and F represents the output feature; S4. Summarize the feature-extracted electrical signal data to construct an SVM model. Simultaneously, train the constructed SVM model using training optimization methods to obtain the final SVM model. Based on the final SVM model, classify the feature-extracted electrical signal data to obtain classified electrical signal data. S5. Construct an electrical signal data hierarchy system based on the characteristics of electrical signals, and perform frequency modulation management on the classified electrical signal data based on the constructed electrical signal data hierarchy system.

2. The method for adjusting the sampling frequency of electromyographic signals according to claim 1, characterized in that, The method involves installing detection electrode devices on the surface of the muscle and collecting electrical signal data generated by the muscle in real time through the installed electrode detection devices. Simultaneously, obtaining the acquired electrical signal data matrix through matrix construction methods includes the following steps: S11. Each acquisition point in the installed electrode detection device includes two sets of electrodes and a differential amplifier; the electrical signals acquired by the two sets of electrodes are set as a set of real-time acquisition of electrical signal data generated by the muscle; after the electrical signal data acquisition is completed, the acquired electrical signal data will be amplified by the differential amplifier; Set the collection time points of the electrode detection equipment , Indicates the set number Each data collection time point This indicates the total number of data collection time points set. S12. Summarize the electrical signal data collected at each set acquisition time point to obtain a first data matrix; and use the first data as the electrical signal data matrix.

3. The method for adjusting the sampling frequency of electromyographic signals according to claim 1, characterized in that, The process of removing noise from the acquired electrical signal data matrix using data denoising to obtain the denoised electrical signal data matrix includes the following steps: S211. Transform the electrical signal data using a wavelet transform function; The wavelet transform function is shown below: ； in, As a scale factor, It is the displacement factor. It is a square-integrable function. For the mother wavelet, It is a wavelet transform function; S212. After the transformation is completed, a threshold is set to perform noise reduction processing on the electrical signal data in the electrical signal data matrix. The noise expression for the electrical signal data in the electrical signal data matrix is ​​shown below: ； in, The electrical signal data in the electrical signal data matrix, Let the standard deviation of the noise be . For noise; The wavelet-transformed electrical signal data is divided by a threshold function, and the electrical signal data that is less than the threshold is adjusted to 0.

4. The method for adjusting the sampling frequency of electromyography signals according to claim 1, characterized in that, The step of filtering the denoised electrical signal data to obtain the filtered electrical signal data matrix includes the following steps: The signal-to-noise ratio is set as the noise reduction evaluation standard, and the noise-reduced electrical signal data is filtered based on the set noise reduction evaluation standard. The signal-to-noise ratio formula is as follows: ； in, Indicates the signal-to-noise ratio. This represents the electrical signal data after noise reduction; The higher the signal-to-noise ratio, the stronger the noise suppression effect. Set a signal-to-noise ratio (SNR) threshold. If the SNR calculated from the denoised electrical signal data is less than the set SNR threshold, the denoising result is unqualified. The denoised electrical signal data is then deleted and filtered. If the SNR calculated from the denoised electrical signal data is greater than or equal to the set SNR threshold, the denoising result is qualified. The filtered electrical signal data are aggregated to form a filtered electrical signal data matrix.

5. The method for adjusting the sampling frequency of electromyography signals according to claim 1, characterized in that, The extracted electrical signal data is used to construct an SVM model. The constructed SVM model is then trained using a training optimization method to obtain the final SVM model. Based on the final SVM model, the extracted electrical signal data is classified to obtain the classified electrical signal data. This process includes the following steps: S41. Construct an initial SVM model; set the kernel function used in the initial SVM model to be a Gaussian kernel function; The mathematical expression for the Support Vector Machine algorithm is shown below: ； Where w represents the weight and b represents the bias value. Indicates sample The classification results In the formula, Represents the Gaussian kernel function; This represents the data in the SVM classification space; This represents the core of the Gaussian kernel function; express and The Euclidean distance between them; Indicates the scope of the Gaussian kernel function; S42. Set the training data ratio as follows: According to the proportion of the training data The extracted electrical signal data is divided into training and testing data matrices; the maximum number of iterations is set to... And the training error threshold is ; The training data matrix is ​​input into the initial SVM model to train the initial SVM model. When the number of training iterations is greater than or equal to... When or the training error is less than When the initial SVM model is stopped, the trained SVM model is obtained. S43. Set a test accuracy threshold, input the test data matrix into the trained SVM model for classification, and obtain the classification result; compare the classification result with the classification label corresponding to the pre-set test data matrix to obtain the classification accuracy; when the classification accuracy is greater than or equal to the test accuracy threshold, use the trained SVM model as the final SVM model; otherwise, repeat S41 and S42 until the classification accuracy is greater than or equal to the test accuracy threshold.

6. The method for adjusting the sampling frequency of electromyography signals according to claim 1, characterized in that, The process of constructing an electrical signal data hierarchy based on the characteristics of electrical signals, and then performing frequency modulation management on the classified electrical signal data based on the constructed electrical signal data hierarchy, includes the following steps: Based on the amplitude range of electromuscular signals, an electrical signal data grading system is established, with each 0.1mV representing one grade. Set a frequency modulation management threshold for electrical signals. When the classified electrical signal data is lower than the set evaluation threshold, increase the sampling frequency of the electrical signal data. When the classified electrical signal data is higher than the set evaluation threshold, decrease the sampling frequency of the electrical signal data.

7. A system for implementing the electromyography signal sampling frequency adjustment method according to any one of claims 1-6, characterized in that, Also includes: Data acquisition equipment, data processing module, data analysis module, feature extraction module, classification module, and frequency modulation management module; The data acquisition device includes a detection electrode device and a differential amplifier, used to acquire and amplify electrical signal data generated by muscles in real time; The data processing module is used to process the real-time acquired electrical signal data to obtain processed electrical signal data. The data analysis module is used to analyze the processed electrical signal data; The feature extraction module is used to extract features from the analyzed electrical signal data; The classification module is used to construct a classification model based on the feature-extracted electrical signal data, and to classify the electrical signal data based on the constructed classification module. The frequency modulation management module is used to construct an electrical signal data hierarchy system, classify the classified electrical signal data into different levels, and perform frequency modulation management based on the classification results.

Images