Electromyographic signal burst activity detection method and electromyographic equipment

Abstract

The invention belongs to the technical field of physiological electric signal detection, and particularly relates to an electromyographic signal burst activity detection method and electromyographic equipment. The electromyographic signal outbreak activity detection method comprises the following steps: constructing a feature matrix based on features of electromyographic signals; extracting principal components in the projection space based on the feature matrix; performing a classification task on the principal components based on a Gaussian mixture model; and quantitative indexes are extracted based on the classification result to represent the electromyographic signal outbreak activity, so that quantitative analysis of signal features is realized, and the robustness of electromyographic activity outbreak detection of gesture recognition is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of physiological electrical signal detection, and in particular relates to a method for detecting myoelectric signal burst activity and an electromyographic device. Background Art

[0002] Electromyographic signals refer to electrical signals that trigger muscle contraction and reflect the degree of muscle contraction. They are widely used in many applied scientific fields such as medicine, rehabilitation, and sports science. In recent years, research on surface electromyographic signals for rehabilitation after stroke and needle electromyographic signals under neurogenic muscle fibers has received increasing attention. The detection of the onset and offset of muscle burst activity under electromyographic signal measurement is not only a key step in its quantification, but also helps to accurately identify muscle activity and improve the real-time performance of electromyographic control applications.

[0003] In the related art, there are generally two methods for detecting the starting point of muscle burst activity. The first is to detect the starting point of EMG excitation based on a single hard threshold of amplitude; but it is sensitive to changes in the background noise level, and as the signal-to-noise ratio of the EMG signal decreases, the false positive rate increases. The second is to detect the EMG starting point based on the training set of EMG signals in the resting state and the burst activity state according to deep learning methods such as RNN; but this method is a supervised detection, which relies on the learning of prior signals on the one hand, and on the other hand, requires the number of classification labels to be set in advance, which reduces the robustness of the model to a certain extent. Since the above-mentioned detection of the starting point of EMG burst activity does not involve quantitative analysis of signal characteristics, the detection results have certain limitations. Summary of the invention

[0004] The present invention provides a method for detecting burst activity of electromyographic signals and an electromyographic device, wherein a characteristic matrix is ​​formed by combining an autoregressive coefficient or a combination of the autoregressive coefficient and other characteristic parameters to extract a principal component for obtaining a quantitative index.

[0005] In order to solve the above technical problems, the present invention provides a method for detecting burst activity of electromyographic signals, comprising: constructing a feature matrix based on characteristic parameters of electromyographic signals; extracting principal components in the projection space based on the feature matrix; performing classification tasks on the principal components based on a Gaussian mixture model; extracting quantitative indicators based on the classification results to characterize the burst activity of electromyographic signals; wherein the characteristic parameters include: an autoregressive coefficient or a combination thereof with at least one of time domain features and frequency domain features.

[0006] Furthermore, the autoregressive coefficient is obtained based on the non-stationary characteristics of the electromyographic signal; the characteristic matrix is ​​an m*n matrix, where m represents the observation quantity and n represents the characteristic dimension.

[0007] Furthermore, the time domain features are extracted based on the amplitude characteristics of the electromyographic signal, including at least one of absolute amplitude energy, waveform length, mean square root, and Hilbert transform envelope energy.

[0008] Furthermore, the frequency domain features are extracted based on multiple sub-band energies of the electromyographic signal under spectrum estimation; the spectrum estimation is configured to perform autoregressive spectrum estimation of the time series after a non-parametric fast Fourier transform or according to an estimated autocorrelation function of the time series to obtain the logarithmic energy of the sub-band.

[0009] Furthermore, extracting the principal components in the projection space based on the feature matrix includes: decentralizing the feature matrix X, that is, X = X- , is the mean of the jth dimension of the sample in the feature matrix; calculate the covariance matrix C of the feature matrix X, that is, C = ; Perform SVD decomposition on the covariance matrix C to obtain the eigenvalue λ and orthogonal eigenvector P in the feature space; project the feature matrix X to a new space based on the matrix P, that is, Y = XP; sort the eigenvalue λ from large to small, and calculate the cumulative variance of each principal component ; When the sum of the contributions of N principal components is greater than the set threshold, the corresponding N principal components are extracted.

[0010] Furthermore, the task of classifying the principal components based on the Gaussian mixture model includes: initializing model parameters; iterating the Gaussian mixture model to converge the model parameters; calculating the probability that the observed sample belongs to each cluster according to the Gaussian mixture model; determining the category of the observed sample according to the probability size, and obtaining the classification result of the principal component.

[0011] Furthermore, the iterative processing includes: calculating the responsiveness of all principal components, that is, the probability that a data point belongs to each Gaussian distribution; updating the parameters of each Gaussian distribution; calculating the likelihood function of the model; and determining whether the model parameters converge, if so, terminating the iterative processing, otherwise, continuing the iterative processing.

[0012] Furthermore, the quantitative indicators include: the starting time and the ending time of a single EMG burst activity, the number of EMG bursts within a certain period of time, and at least one of the average EMG burst time and burst intensity within a certain period of time.

[0013] In a second aspect, the present invention provides an electromyography device, comprising: an acquisition module, which collects different electromyographic signals for different gestures; a processor, which implements the steps of the detection method when executed, and uses the quantitative indicators of the electromyographic signals as parameters of the clustering model to achieve online detection; and a result output module, which outputs the quantitative indicators to characterize muscle burst activity.

[0014] In a third aspect, the present invention provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method when executed by a processor.

[0016] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which implements the steps of the method when executed by a processor.

[0017] The beneficial effect of the present invention is that the detection method and electromyography device of the electromyographic signal burst activity of the present invention use the autoregressive coefficient or the combination of the autoregressive coefficient and at least one of the time domain feature and the frequency domain feature as the feature parameter to construct a feature matrix, and then select the principal components of the projection space according to the square energy to improve the signal-to-noise ratio of the signal; for the principal components under the selected projection space, the maximum likelihood estimation method is used to continuously iteratively calculate the parameters of each Gaussian distribution to obtain a Gaussian mixture model; the probability that the selected principal component belongs to each Gaussian distribution is calculated according to the Gaussian mixture model to realize unsupervised soft classification, and quantitative indicators are extracted according to the classification results to characterize the electromyographic signal burst activity, thereby realizing the quantitative analysis of the signal characteristics and improving the robustness of the electromyographic activity burst detection of gesture recognition.

[0018] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 The operating procedure of the detection method is shown.

[0022] Figure 2 shows the results of data preprocessing.

[0023] Figure 3 The operation flow of the detection method that uses autoregressive coefficients, time domain features, and frequency domain features to form a feature matrix is ​​shown.

[0024] Figure 4 The calculation results of the cumulative variance of the principal components are shown.

[0025] Figure 5 Gaussian classification results based on the characteristics of the autoregressive model parameters estimated by Burg's method are shown.

[0026] Figure 6 The Gaussian classification results of the first two principal components extracted in the projected space are shown.

[0027] Figure 7 The results show the detection results of the starting and ending moments of the EMG signal of the flexor digitorum profundus when the left hand makes a fist.

[0028] Figure 8 The detection results of electromyographic burst activity based on the feature matrix constructed based on time domain features are shown.

[0029] Fig. 9 The results of detecting EMG burst activity based on the feature matrix composed of autoregressive coefficients and frequency domain features are shown.

[0030] Fig.10 for Fig. 9 A partial enlarged view of point A in the middle.

[0031] Fig.11 The results of detecting EMG burst activity based on the feature matrix constructed by autoregressive coefficients are shown.

[0032] Fig.12 The results of detecting EMG burst activity based on the feature matrix composed of autoregressive coefficients and time domain features are shown.

[0033] Fig.13 The detection results of electromyographic burst activity based on the feature matrix composed of autoregressive coefficients, time domain features, and frequency domain features are shown. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] Example 1.

[0036] See Figure 1-Figure 7 Embodiment 1 provides a method for detecting burst activity of electromyographic signals, comprising the following steps: step S1, preprocessing the original signal; step S2, extracting features from the processed signal to obtain a feature matrix, and performing orthogonal transformation, normalization or standardization on the feature matrix; step S3, selecting a projection space composed of principal components whose variance cumulative energy is greater than a threshold; step S4, spatially clustering the principal components under the selected projection space; step S5, extracting quantitative indicators based on the classification results to characterize the burst activity of electromyographic signals, such as but not limited to using quantitative parameters such as the starting point, offset, and end point of the electromyographic burst activity as parameters of the clustering model; step S6, continuously updating and iterating the clustering model during the online prediction process. This detection method can be used to detect the starting point and offset of the electromyographic burst of surface electromyographic signals / needle electromyographic signals, and realize quantitative evaluation of muscle activity, such as the duration of each burst, the intensity of the burst, etc.; it can be used for electromyographic activity burst detection in gesture recognition, and is a key method for gesture classification based on electromyographic signals.

[0037] Optionally, the original signal may be a unit discharge recorded extracellularly, a single-channel / multi-channel intramuscular electromyographic signal, a surface electromyographic signal, etc.

[0038] Step S1, data preprocessing.

[0039] The original signal is preprocessed before extracting characteristic parameters. The data preprocessing includes: removing noise signals, such as but not limited to time domain filtering, frequency domain filtering and wavelet decomposition. A filter is designed to remove noise signals, such as 50 Hz power frequency interference, sweating and other low-frequency noise. The filter includes but is not limited to a high-pass finite impulse response FIR filter, such as a 0.5 Hz FIR high-pass filter, whose amplitude characteristic curve and phase characteristic curve are as follows: Figure 2 As shown; for example, a notch filter of 50 Hz can realize a notch filter with a cutoff frequency of 48~52h. The EMG signal with noise interference removed is subjected to sliding window slicing, that is, the processed offline EMG signal is subjected to extraction of EMG signal segments of a specific window length according to a certain time step, wherein the time length of the window is, for example, 0.2 seconds, and the time step is, for example, 0.1 seconds.

[0040] Step S2, constructing a feature matrix.

[0041] Feature parameters are extracted from the EMG signal segments, and a feature matrix is ​​constructed based on the feature parameters of the EMG signal; wherein the feature parameters include: autoregressive coefficients or a combination thereof with at least one of time domain features and frequency domain features. Specifically, the feature matrix is ​​an m*n matrix, where m represents the observed quantity and n represents the feature dimension. When the feature parameters are only autoregressive coefficients, the feature matrix is ​​an m*1 matrix; when the feature parameters are autoregressive coefficients and time domain features or frequency domain features, the feature matrix is ​​an m*2 matrix; when the feature parameters are autoregressive coefficients and time domain features or frequency domain features, the feature matrix is ​​an m*3 matrix. The detection process of EMG signal burst activity is as follows: Figure 3 shown.

[0042] Optionally, the autoregressive coefficient is obtained based on the non-stationary characteristics of the electromyographic signal. , further strengthen the feature description of the electromyographic signal. For example, select the third-order autoregressive model and obtain the autoregressive coefficients a1, a2, and a3.

[0043] Optionally, the time domain features are extracted based on the amplitude characteristics of the electromyographic signal, such as absolute amplitude energy , waveform length , mean square root , Hilbert transform envelope energy, etc.

[0044] Optionally, the frequency domain features are extracted from multiple sub-band energies of electromyographic signals under spectrum estimation. Since electromyographic activity is a high-frequency electrical activity in a wide frequency band, the frequency domain features include M sub-band energy extractions under spectrum estimation (not limited to parametric and non-parametric spectrum estimation), such as after non-parametric fast Fourier transform or according to the estimated autocorrelation function of the time series to perform autoregressive spectrum estimation of the time series, to achieve the logarithmic energy of the sub-bands. , is the imaginary value after fast Fourier transform in each sub-band. Represents the frequency after fast Fourier transform. For example, select four sub-bands: 0-45.25hz, 45.25-90.5hz, 90.5hz-135.75hz, 135.75hz-181hz.

[0045] Optionally, the normalization process is, for example but not limited to, min-max normalization. The standardization process is for example but not limited to Z-score standardization. .

[0046] Optionally, the orthogonal space transformation may be generalized eigenvalue decomposition, singular value decomposition, etc.

[0047] Step S3, selecting a projection space.

[0048] Based on the principal component analysis method, the projection space consisting of principal components with cumulative variance energy greater than the threshold is selected. Principal component analysis (PCA) is a feature extraction and dimensionality reduction algorithm. The goal of PCA is to find a set of ordered orthogonal basis vectors that can capture the maximum variation in the data. The basis vectors are called principal components (PC), which include: Step S31, set the sample matrix X, whose dimension is m*n, indicating that there are m samples of n dimension, and decentralize the sample matrix, X = X- , is the mean of the j-th dimension of the samples.

[0049] Step S32, calculate the covariance matrix C of X, C = .

[0050] Step S33, perform SVD decomposition on the covariance matrix C to obtain the eigenvalue λ and orthogonal eigenvector P in the feature space, and project the sample matrix X into a new space based on the P matrix, Y = XP.

[0051] Step S34, sorting the eigenvalues ​​λ from large to small, wherein the larger the λ, the greater the difference in the principal components in the corresponding projection space.

[0052] Step S35, calculate the cumulative variance of each principal component ;See Figure 4 , when the sum of the contributions of N principal components is greater than the set threshold thd_c, the corresponding N principal components can be extracted based on the projection variance, eigenvalue, etc. of the basis vector. Figure 4 In the example, the threshold value thd_c is set to be, for example but not limited to, 95%.

[0053] Step S4: spatial clustering.

[0054] Based on the Gaussian mixture model, the principal components in the selected projection space are classified. The spatial clustering method can be a partition clustering method, a density clustering method, a hierarchical clustering method, etc. The Gaussian mixture model algorithm assumes that the sample points are generated by one or more Gaussian distributions, and estimates the parameters of the Gaussian distribution of each cluster by the maximum likelihood estimation method. The steps of solving the Gaussian mixture model parameters using the expectation maximization algorithm include: Step S41, initializing the parameter model, such as the mean and covariance of each Gaussian distribution; in order to predict whether there is electromyographic burst activity, two mixed multi-dimensional Gaussian distributions can be designed.

[0055] Step S42, calculating the responsiveness of all data points, that is, the probability that the data point belongs to each Gaussian distribution, wherein the data points are the principal components in the extracted projection space.

[0056] Step S43, updating the parameters of each Gaussian distribution.

[0057] Step S44, calculating the likelihood function of the model.

[0058] Step S45, iteratively determine whether the parameters of the model have converged. If converged, the iteration ends; otherwise, return to step S42.

[0059] Step S46, according to the Gaussian mixture model, the probability that the observed sample belongs to each cluster is calculated to complete the soft classification.

[0060] Step S47, calculating the probability that the observed sample belongs to each cluster.

[0061] Step S48, determine the category of the observed sample according to the probability size, and obtain the classification result of the principal component.

[0062] like Figure 5 As shown, in the original feature space, the first eigenvector is set as the autoregressive model parameter estimated based on the Burg method, expressed as Figure 5 In Figure 1, the second eigenvector is the autoregressive model parameter estimated based on the Burg method, expressed as Figure 5 Figure 2 in Figure 2, from Figure 5 It can be seen from the figure that in the original feature space, the distances between the two cluster centers of the binary Gaussian after Gaussian classification are relatively close. Figure 6 As shown, the Gaussian mixture model is unsupervisedly classified in the projection space to extract the corresponding principal components, namely the first principal component and the second principal component. Figure 6 It can be seen that the principal components extracted in the projection space effectively remove the redundancy of the variables in the feature space and improve the classification accuracy of the mixed Gaussian model. In summary: (1) For the principal components extracted in the projection space, the distance between each cluster center of the binary Gaussian after unsupervised training is farther and the probability of overlap is smaller; (2) In the original feature space, the distance between each cluster center of the binary Gaussian after unsupervised training is relatively close; (3) The principal components extracted in the projection space effectively remove the redundancy of the variables in the feature space and improve the classification accuracy of the mixed Gaussian model.

[0063] Step S5, extracting quantitative indicators.

[0064] According to whether there is a burst after classification by Gaussian mixture model, the starting point, offset, end point and other quantitative parameters of the EMG burst activity are extracted. The quantitative indicators include but are not limited to: (1) the starting time of a single EMG burst activity Start_i; (2) the ending time of a single EMG burst activity Stop_i; (3) the number of EMG bursts of a certain duration, such as the number of bursts per minute; (4) the average EMG burst time and burst intensity under a certain duration; (5) statistical calculation of burst time and burst intensity, etc. For example, see Figure 7 , while making a fist with the left hand, Figure 7 The electromyographic signal in the figure is the electromyographic signal of the deep flexor muscle of the forearm. Figure 7 The Gaussian model classification in the figure is the result of detecting the EMG burst activity of the EMG signal based on the Gaussian mixture model, where a high level 1 represents the detection of muscle burst activity, and a low level 0 represents the absence of muscle activity. When it jumps from 0 to 1, it is recorded as the starting point of the burst 'start', and when it changes from 1 to 0, it is recorded as the end of the burst 'stop'. Calculate statistics, such as the number of bursts per minute, to extract quantitative parameters, such as the starting point, the length of a single burst, etc.

[0065] Example 2.

[0066] Based on Example 1, this Example 2 provides an electromyography device, including: an acquisition module, which collects different electromyographic signals for different gestures; a processor, which implements the steps of the detection method when executed, obtains quantitative indicators of the electromyographic signals, and implements online electromyographic burst detection and quantification based on the clustering model; and a result output module, which outputs the quantitative indicators to characterize muscle burst activity.

[0067] Example 3.

[0068] Based on the first embodiment, the third embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0069] Example 4.

[0070] Based on Example 1, this Example 4 provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented.

[0071] Example 5.

[0072] Based on Embodiment 1, this Embodiment 5 provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the method are implemented.

[0073] Test example.

[0074] Experimental condition 1: The feature matrix is ​​constructed only through time domain features, and the hard threshold is used to detect the burst activity of the electromyographic signal based on the statistics of the time domain amplitude features. The detection results are as follows: Figure 8 As shown, Figure 8 The blue curve in the middle represents the result of taking the absolute value of the electromyographic signal, the red curve is the result of calculating the envelope of the signal by Hilbert transform, and the black straight line represents the hard threshold calculated based on the statistical average. Obviously, the detection results under experimental condition 1 are sensitive to time domain signal noise, and the time domain energy during electromyographic burst activity is not stable, resulting in large errors in the starting point and offset of the detection burst.

[0075] Experimental condition 2: The feature matrix is ​​constructed by using autoregressive coefficients and frequency domain features, and the burst activity of electromyographic signals is detected based on the mixed Gaussian model. The detection results are as follows: Fig. 9 and Fig.10 shown. Fig. 9 The EMG signal in the figure shows the electrical signal of the muscle activity of the extensor muscle of the little finger. It can be seen that there is slight EMG activity between the time intervals of two muscle bursts. Fig. 9 The Gaussian model classification in Figure 1 shows the detection results of EMG burst activity based on the construction of feature matrix based on autoregressive coefficients and frequency domain features and the mixed Gaussian model. Fig.10 Shown Fig. 9 The enlarged image of the detection results in the time period of 11.5 seconds to 15 seconds shows that this feature combination has high robustness and can effectively reduce the classification error of mild muscle activity.

[0076] Experimental condition 3: The feature matrix is ​​constructed only by the autoregressive coefficients, and the burst activity of the electromyographic signal is detected based on the mixed Gaussian model. The detection results are as follows: Fig.11 shown. Fig.11 The electromyographic signals in the figure show the electrical signals of the muscle activity of the extensor digitorum digiti minimi. Fig.11 The Gaussian model classification in the figure shows the detection results of EMG burst activity based only on the autoregressive coefficient feature matrix and the mixed Gaussian model. The results show that there are certain errors in the classification results that rely solely on the autoregressive coefficient features.

[0077] Experimental condition 4: The feature matrix is ​​constructed by using autoregressive coefficients and time domain features, and the burst activity of electromyographic signals is detected based on the mixed Gaussian model. The detection results are as follows: Fig.12 shown. Fig.12 The EMG signal in the figure shows the electrical signal of muscle activity of the extensor digiti minimi muscle; Fig.12 The Gaussian model classification in the figure shows the detection results of the burst activity of the electromyographic signal detected by the mixed Gaussian model based on the feature matrix constructed based on the autoregressive coefficient and the time domain features. The results show that the classification results of this feature combination have a small amount of error.

[0078] Experimental condition 5: The feature matrix is ​​constructed by autoregressive coefficients, time domain features, and frequency domain features, and the burst activity of electromyographic signals is detected based on the mixed Gaussian model. The detection results are as follows: Fig.13 shown. Fig.13 The EMG signal in the figure shows the electrical signal of muscle activity of the extensor digiti minimi muscle; Fig.13 The Gaussian model classification in the figure shows the construction of a feature matrix based on autoregressive coefficients, time domain and frequency domain features, and the detection results of EMG burst activity detected by the mixed Gaussian model. The results show that this feature combination has high robustness.

[0079] Test results: The test results under different test conditions are compared, as shown in Table 1. Obviously, the test results of test conditions 2-5 are significantly better than those of test condition 1, that is, the detection method in this case effectively captures the template characteristics of EMG burst activity by using the Gaussian mixture model of the principal components extracted in the projection space, and has strong robustness to noise interference signals and weak EMG activities; the Gaussian mixture model is continuously updated and iterated during the online prediction process, which can effectively improve the stability of the convergence of the multivariate mixture Gaussian parameters.

[0080] Table 1 Test results under different test conditions.

[0081] Test conditions Test results Test condition 1 The traditional hard threshold detection method that only relies on time domain features has jumpy detection of continuous muscle burst activity and cannot effectively quantify muscle burst starting point, burst duration and other characteristics. Test condition 2 The mixed Gaussian classification based on the combination of autoregressive coefficients and frequency domain features has high robustness. On the one hand, it can effectively reduce the classification error of mild muscle activity. On the other hand, it can effectively detect continuous muscle burst activities and realize the quantification of muscle burst starting point, burst duration, burst number and other characteristics. Test condition 3 The classification results based only on the autoregressive coefficient feature have certain classification errors and cannot effectively capture the characteristics of continuous EMG bursts, which has a certain impact on the quantification of the starting point and number of EMG bursts in the later stage. Test condition 4 The mixed Gaussian classification based on the combination of autoregressive coefficients and time domain features has a certain degree of robustness and is sensitive to the starting features. It can basically achieve the quantification of muscle burst starting point, burst duration, and other features. Test condition 5 The mixed Gaussian classification based on the combination of autoregressive coefficients, time domain and frequency domain features has high robustness. On the one hand, it can effectively reduce the classification error of mild muscle activity. On the other hand, it can effectively detect continuous muscle burst activities and realize the quantification of muscle burst starting point, burst duration, burst number and other characteristics. In the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0082] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0083] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0084] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A method for detecting burst activity of electromyographic signals, characterized in that: include: Construct a feature matrix based on the feature parameters of the electromyographic signal; Extract the principal components in the projection space based on the feature matrix; Classification task of principal components based on Gaussian mixture model; Extract quantitative indicators based on the classification results to characterize the burst activity of EMG signals; in The characteristic parameters include: an autoregressive coefficient or a combination thereof with at least one of a time domain feature and a frequency domain feature.

2. The detection method according to claim 1, characterized in that: The autoregressive coefficient is obtained based on the non-stationary characteristics of the electromyographic signal; The feature matrix is ​​an m*n matrix, where m represents the observation quantity and n represents the feature dimension.

3. The detection method according to claim 2, characterized in that: The time domain features are extracted based on the amplitude characteristics of the electromyographic signal, including at least one of absolute amplitude energy, waveform length, mean square root, and Hilbert transform envelope energy.

4. The detection method according to claim 2, characterized in that: The frequency domain feature is extracted based on multiple sub-band energies of the electromyographic signal under spectrum estimation; The spectrum estimation is configured to perform autoregressive spectrum estimation of the time series after non-parametric fast Fourier transformation or according to an estimated autocorrelation function of the time series to obtain the logarithmic energy of the sub-band.

5. The detection method according to claim 1, characterized in that: Extracting the principal components in the projection space based on the feature matrix includes: Decentralize the feature matrix X, that is, X = X- , is the mean of the j-th dimension of the samples in the feature matrix; Calculate the covariance matrix C of the feature matrix X, that is, C = ; Perform SVD decomposition on the covariance matrix C to obtain the eigenvalue λ and orthogonal eigenvector P in the feature space; Project the feature matrix X to a new space based on the matrix P, that is, Y = XP; Sort the eigenvalues ​​λ from large to small and calculate the cumulative variance of each principal component ; When the sum of the contributions of N principal components is greater than the set threshold, the corresponding number of principal components is extracted.

6. The detection method according to claim 1, characterized in that: The tasks of classifying principal components based on Gaussian mixture models include: Initialize model parameters; Iterate the Gaussian mixture model to converge the model parameters; Calculate the probability that the observed sample belongs to each cluster according to the Gaussian mixture model; Determine the category of the observed sample according to the probability size and obtain the classification result of the principal component.

7. The detection method according to claim 6, characterized in that: The iterative process includes: Calculate the responsiveness of all principal components, i.e. the probability that a data point belongs to each Gaussian distribution; Update the parameters of each Gaussian distribution; Calculate the likelihood function of the model; Determine whether the model parameters have converged. If so, end the iterative process; otherwise, continue the iterative process.

8. The detection method according to claim 1, characterized in that: The quantitative indicators include: the starting point and ending point of a single EMG burst activity, the number of EMG bursts within a certain period of time, and at least one of the average EMG burst time and burst intensity within a certain period of time.

9. An electromyography device, characterized in that: include: The acquisition module collects different electromyographic signals for different gestures; A processor, which when executed implements the steps of the detection method according to any one of claims 1 to 8, and uses the quantitative index of the electromyographic signal as a parameter of the clustering model to achieve online detection; The result output module outputs the quantitative index to characterize the muscle burst activity.

10. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the detection method according to any one of claims 1 to 8.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the detection method according to any one of claims 1 to 8 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the detection method according to any one of claims 1 to 8 are implemented.

Images