Transfer entropy coupling analysis method based on multi-channel surface electromyography signal decomposition

Abstract

The present invention discloses a transfer entropy coupling analysis method based on the decomposition of multi-channel surface electromyography signals. The present invention first uses a convolution kernel compensation method to decompose the multi-channel surface electromyography signals into motor units, and decomposes the electromyography signals originally superimposed by multiple motor units. After the motor units are decomposed, the transfer entropy is used to establish a correlation matrix, and the threshold method or the fixed weighted edge method is used to remove the weaker connection edges, and the intermuscular network model related to the motor function is constructed. The undirected graph of each frequency band of the electromyography is drawn, and the intermuscular network indicators such as the connectivity rate and the small world characteristics are calculated to establish a complex network, revealing the functional connection and mutual influence between the limb motor areas. Through the decomposition of the motor units, the transfer entropy is calculated to construct a complex network to fundamentally analyze the coupling characteristics of the electromyography signals. A new method is provided for studying intermuscular coupling analysis.

Description

Technical Field

[0001] The present invention belongs to the field of biological signal processing, and relates to analyzing coupling effects between multi-channel surface electromyographic signals by using a complex network method after decomposing motion units of multi-channel surface electromyographic signals and then calculating transfer entropy. Background Art

[0002] The smallest unit of muscle is the motor unit, which is composed of α neurons and the muscle fibers it controls. Surface electromyography is the superposition of action potentials of motor units in many muscle fibers in time and space. The essence of electromyography signal decomposition is to restore the signal generated by the smallest muscle unit, the motor unit, by processing the superimposed signal to obtain the most essential release information of the muscle. Analysis of the motor unit waveform obtained after decomposition can provide a basis for the diagnosis of muscle diseases in clinical practice, and the recruitment and release information of the motor unit can be used to help diagnose the function of the central nervous system. The decomposition of surface electromyography can better explain the behavioral relationship between different motor units and the collaborative working characteristics of motor nerves between different muscles. In general, the decomposition of surface electromyography has great practical application value for analyzing the function of the neuromuscular system. In view of these advantages, in recent years, the research on multi-channel surface electromyography signal decomposition has gradually increased. Zazula used the minimum mean square error optimization method to decompose the signal. Holobar first proposed the blind source separation method to decompose electromyography signals, and a few years later proposed the convolution kernel compensation method to greatly improve the decomposition performance. De Luca established an artificial intelligence framework to decompose the inserted EMG signals and used needle electrode insertion measurements to verify the accuracy of the decomposition.

[0003] The concept of intermuscular coupling comes from the study of cortical muscle functional coupling. The coupling phenomenon between muscles can reflect the interaction between muscles during the movement process and the control ability of the central nervous system to muscles. Because the motor neurons that control the relevant muscle activities share consistent cortical spinal cord contributions under specific motor tasks, this unique coupling relationship is formed. Intermuscular coupling not only reflects the motor function state of the muscle, but also includes the motor control instructions of the central nervous system to the muscle. Studying the coupling information between surface electromyographic signals can not only reflect the motor control mechanism of the central nervous system, but also realize the decoding of electromyographic signals and correctly interpret the muscle movement control mechanism. In recent years, wavelet decomposition, modal decomposition, Pearson correlation coefficient, transfer entropy, Granger causality, coherence and other methods have been applied to intermuscular coupling analysis. Among them, transfer entropy is used to study the transfer of information between variables. It is a method based on probability distribution, Shannon entropy and statistical methods to derive causality between time series. The value of this information transfer that can reduce the uncertainty of the observed system can be calculated. When the transfer entropy of X to Y is greater than the transfer entropy of Y to X, we call X the cause and Y the effect, and use this to establish a causal relationship between the two variables.

[0004] Traditional muscle coupling analysis only uses undecomposed electromyographic signals, that is, signals after the potentials of many motor units are superimposed, which has certain limitations. Traditional coupling analysis methods cannot effectively describe the characteristics of muscle coupling. At the same time, traditional coupling does not trace back to the most essential motor unit of the muscle. The measured coupling relationship includes the relationship after the superposition of multiple motor units, and the coupling strength between muscles is not accurate enough. At the same time, the development of complex network theory has laid a foundation for studying the processing and flow of information in functional networks, and can realize the coupling analysis between multi-channel electromyographic signals. The muscle function network reflects the information transmission and division of labor and cooperation mechanism of each muscle during limb movement from a global perspective, which is an effective way to explore the laws of sensory muscle activity. In response to these problems, a method is proposed to first decompose the multi-channel surface electromyographic signals into motor units, and then establish a complex network through the transfer entropy of the decomposed motor unit signals to analyze the coupling effect between multi-channel electromyographic signals. Summary of the invention

[0005] The purpose of the present invention is to provide a transfer entropy coupling analysis method based on multi-channel surface electromyography signal decomposition

[0006] To achieve the above object, the method of the present invention mainly comprises the following steps:

[0007] Step 1: synchronously collect and preprocess multi-channel surface electromyography signals;

[0008] Step 2: Decompose the preprocessed electromyographic signal into motor units and classify the resulting firing sequences;

[0009] Step 3: After obtaining the best series of firing sequences in step 2, the cross-correlation coefficient between the decomposed motor unit signal and the electromyographic signal of each channel is calculated based on the similarity of the firing sequence of the same motor unit in amplitude and firing interval, and the electromyographic signal of each channel is reconstructed;

[0010] Step 4: Use a bandpass filter to perform frequency segmentation processing on the electromyographic signal reconstructed in step 3;

[0011] Step 5: For the multi-channel electromyographic signals in each frequency band in step 4, the transfer entropy quantization method is used to calculate the connection value between each channel electromyographic signal and establish a correlation matrix;

[0012] Step 6: Based on the correlation matrix obtained in step 5, a complex network method is used to analyze the coupling effects between multi-channel EMG signals. Specifically, the threshold method or fixed weighted edge method is used to remove weaker connection edges, an intermuscular network model related to motor function is constructed, an undirected graph of each EMG frequency band is drawn, the connectivity rate and small-world characteristics are calculated, and the functional connections and mutual influences between limb motor areas are revealed.

[0013] Preferably, step 2: decomposing the preprocessed electromyographic signal into motor units, and classifying and arranging the obtained release sequences;

[0014] The specific steps include:

[0015] Step 2.1: Calculate the cross-correlation matrix C of the preprocessed multi-channel surface electromyography signal XX and the inverse of the cross-correlation matrix, where C XX =E(X(n)X T (n)), X(n) is the signal of the nth channel, T is the matrix transpose, and E is the variance;

[0016] Step 2.2: Calculate the energy released by the motor unit in each channel using equation (1):

[0017]

[0018] Among them, y(n) is the energy of the motion unit, and -1 is the inverse of the matrix;

[0019] Step 2.3: Select the initial time n0 as the time corresponding to the median value of y(n), and obtain the initial release sequence S through formula (2): n0 ;

[0020]

[0021] Find S n0 The maximum peak value in the sequence S is recorded as n1, and the time when the peak value appears is recorded as n1. Then, the sequence S is calculated using formula (2) n1 , from the sequence Sn1 Find the time corresponding to the k largest peaks, recorded as

[0022] Step 2.4: Calculate C using formula (3)

[0023]

[0024] in, Representing a collection The number of elements in; replace X(n0) in formula (2) with C to reconstruct the final release sequence

[0025] Step 2.5: Setup

[0026] Step 2.6: Repeat steps 2.3-2.5 multiple times to ensure that as many release sequences as possible can be reconstructed, and classify and organize the obtained release sequences.

[0027] Preferably, in step 4, the signal is divided into an alpha (8-15 Hz) frequency band, a beta (15-30 Hz) frequency band, and a gamma (30-60 Hz) frequency band.

[0028] Preferably, in step 5, the transfer entropy is calculated as follows:

[0029] Two time series X = {x1, x2…, x n}, Y={y1,y2…,y n}, where n is the length of the time series, x i ,y i are the i-th observation values ​​respectively, then the transfer entropy from X to Y in biomedical signal analysis is expressed as

[0030]

[0031] p(·) represents the probability value; where τ is the time lag of X.

[0032] Compared with the traditional muscle coupling analysis method, the present invention has the following advantages:

[0033] Traditional intermuscular coupling analysis methods cannot accurately and fundamentally describe the coupling characteristics of surface electromyographic signals. To address this problem, the present invention proposes a new coupling analysis method. The method first uses the convolution kernel compensation method to decompose the multi-channel surface electromyographic signals into motor units, and decomposes the electromyographic signals that were originally superimposed by multiple motor units. After the motor unit is decomposed, the transfer entropy is used to establish a correlation matrix, and the threshold method or the fixed weighted edge method is used to remove the weaker connection edges, and the intermuscular network model related to the motor function is constructed. The undirected graph of each frequency band of electromyography is drawn, and the intermuscular network indicators such as connectivity rate and small-world characteristics are calculated to establish a complex network, revealing the functional connection and mutual influence between the limb motor areas. Through motor unit decomposition, the transfer entropy is calculated to construct a complex network to fundamentally analyze the coupling characteristics of the electromyographic signals. Provide a new method for studying intermuscular coupling analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Flow chart of the method of the present invention.

[0035] Figure 2 (a) and Figure 2 (b) are the EMG signals of channel 1 and channel 2 after preprocessing.

[0036] Figure 3 It is the motor unit release sequence after the multi-channel surface electromyography signal is decomposed.

[0037] Figure 4 It is the transfer entropy between 1 channel and 2 channels and the transfer entropy between 2 channels and 1 channel.

[0038] Figure 5 After decomposition, the transfer entropy between 1 channel and 2 channels and the transfer entropy between 2 channels and 1 channel. DETAILED DESCRIPTION

[0039] The following is a detailed description of the embodiments of the present invention in conjunction with the accompanying drawings: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and specific operation process are given. Figure 1 As shown, this example includes the following steps

[0040] Step 1: Multi-channel surface electromyography signal acquisition and preprocessing, such as Figure 2 As shown in (a) and (b).

[0041] Step 2: Decompose the preprocessed data into motion units using the convolution kernel compensation algorithm.

[0042] Considering the neuromuscular system of electromyographic signals as a linear time-invariant multi-input multi-output system, the data model of the convolution kernel method can be expressed by the following equation:

[0043]

[0044] Where X(n) is the observed signal of m channels, s(n) is the firing sequence of n motor units, e(n) is zero-mean noise, and H is the mixing matrix, i.e., the convolution kernel. The core idea of ​​the convolution kernel compensation algorithm is not to calculate the convolution kernel H, but to directly estimate the firing sequence s(n) of the motor unit from the observed multi-channel surface electromyography signal X(n). It specifically includes the following steps:

[0045] Step 2.1: Delay the sEMG signal X(n) of m channels K-1 times to enhance the decomposition effect of the signal and obtain the expanded form of the signal Then through Calculate the release energy of the motor unit in the signal, where C XX =E(X(n)X T (n))

[0046] Step 2.2: Select the initial time n0 as the time corresponding to the median value of y(n), through Get the initial release sequence S n0 If only the i-th motor unit is released at time n0, S n0 (n) to estimate

[0047] Step 2.3: To increase the accuracy of the decomposition, find S n0 The maximum peak value in the equation is recorded as n1. Calculate the sequence S n1 , from the sequence S n1 Find the time corresponding to the k largest peaks, recorded as

[0048] Step 2.4: Use X(n1) Instead, Representing a collection The number of elements in . Then the final reconstructed release sequence can be expressed as

[0049] Step 2.5: Setup

[0050] Step 2.6: Repeat steps 2.2-2.5 multiple times to ensure that as many firing sequences as possible can be reconstructed, and classify and sort the resulting firing sequences. Figure 3 shown.

[0051] Step 3: After obtaining the optimal series of firing sequences in step 2, the firing sequences of the same motor unit are similar in amplitude and firing interval, and the cross-correlation coefficient between the decomposed motor unit signal and the electromyographic signal of each channel is calculated to reconstruct the electromyographic signal of each channel;

[0052] Step 4: Use a bandpass filter to perform frequency band processing on the electromyographic signal reconstructed in step 3, and divide the signal into alpha (8-15 Hz), beta (15-30 Hz), and gamma (30-60 Hz) bands;

[0053] Intermuscle coupling seems to have significant frequency band characteristics, which are mainly manifested in the low-frequency band: the intermuscle coupling in the alpha (8-15Hz) band is driven by the spinal nerves, which has a certain influence on the involuntary contraction of muscles and the fixation of movement posture; the intermuscle coupling in the beta (15-30Hz) band expresses the oscillation and integration process of neural information from the primary motor cortex to the outgoing neurons, which is related to the output of static force; the intermuscle coupling in the gamma (30-60Hz) band is controlled by the cortical nerves, which is manifested as the tense contraction caused by concentration in the cognitive process.

[0054] Step 5: For the multi-channel electromyographic signals in each frequency band in step 4, the transfer entropy quantization method is used to calculate the connection value between each channel electromyographic signal and establish a correlation matrix.

[0055] Entropy is a measure of uncertainty. The greater the amount of information, the smaller the uncertainty, the smaller the entropy; the smaller the amount of information, the greater the uncertainty, the greater the entropy. Transfer entropy is an effective measure of information certainty and can be used to establish a causal relationship between two variables. The calculation method of transfer entropy is as follows:

[0056] Two time series X = {x1, x2…, x n}, Y={y1,y2…,y n}, where n is the length of the time series, x i ,y i are the i-th observation values ​​respectively, then the transfer entropy from X to Y can be calculated by the transfer entropy

[0057]

[0058] Where τ and t are the time lags of X and Y, respectively, and k and l are the block lengths of past values ​​of X and Y, respectively. In biomedical signal processing, the parameters are usually set to k = l = t = 1, so the above equation can be simplified to

[0059]

[0060] Step 6: Based on the correlation matrix obtained in step 5, a complex network method is used to analyze the coupling effects between multi-channel EMG signals. Specifically, the threshold method or fixed weighted edge method is used to remove weaker connection edges, an intermuscular network model related to motor function is constructed, an undirected graph of each EMG frequency band is drawn, and intermuscular network indicators such as connectivity and small-world characteristics are calculated to reveal the functional connections and mutual influences between limb motor areas.

[0061] The transfer entropy between 1 channel and 2 channels and the transfer entropy between 2 channels and 1 channel before decomposition are as follows: Figure 4 As shown. After decomposition, the transfer entropy between 1 channel and 2 channels and the transfer entropy between 2 channels and 1 channel are as follows Figure 5 By comparison Figure 4 Figure 5 It can be seen that after the motor unit is decomposed and reconstructed, the transfer entropy in the alpha band (8-15Hz), beta band (15-30Hz), and gamma band (30-60Hz) is improved, which verifies that the decomposition of the surface electromyographic signal can more accurately describe the inter-muscle coupling relationship.

Claims

1. A transfer entropy coupling analysis method based on multi-channel surface electromyography signal decomposition, characterized by: The method comprises the following steps: Step 1: synchronously collect and preprocess multi-channel surface electromyography signals; Step 2: Decompose the preprocessed electromyographic signal into motor units and classify the resulting firing sequences; Step 3: After obtaining the best series of firing sequences in step 2, the cross-correlation coefficient between the decomposed motor unit signal and the electromyographic signal of each channel is calculated based on the similarity of the firing sequence of the same motor unit in amplitude and firing interval, and the electromyographic signal of each channel is reconstructed; Step 4: Use a bandpass filter to perform frequency segmentation processing on the electromyographic signal reconstructed in step 3; Step 5: For the multi-channel electromyographic signals in each frequency band in step 4, the transfer entropy quantization method is used to calculate the connection value between each channel electromyographic signal and establish a correlation matrix; Step 6: Based on the correlation matrix obtained in step 5, a complex network method is used to analyze the coupling effects between multi-channel EMG signals. Specifically, the threshold method or fixed weighted edge method is used to remove weaker connection edges, an intermuscular network model related to motor function is constructed, an undirected graph of each EMG frequency band is drawn, the connectivity rate and small-world characteristics are calculated, and the functional connections and mutual influences between limb motor areas are revealed.

2. The transfer entropy coupling analysis method based on multi-channel surface electromyography signal decomposition according to claim 1 is characterized in that: Step 2: Decompose the preprocessed electromyographic signal into motor units and classify the resulting firing sequences; The specific steps include: Step 2.1: Calculate the cross-correlation matrix C of the preprocessed multi-channel surface electromyography signal XX and the inverse of the cross-correlation matrix, where C XX =E(X(n)X T (n)), X(n) is the signal of the nth channel, T is the matrix transpose, and E is the variance; Step 2.2: Calculate the energy released by the motor unit in each channel using equation (1): Among them, y(n) is the energy of the motion unit, and -1 is the inverse of the matrix; Step 2.3: Select the initial time n0 as the time corresponding to the median value of y(n), and obtain the initial release sequence S through formula (2): n0 ; Find S n0 The maximum peak value in the sequence S is recorded as n1, and the time when the peak value appears is recorded as n1. Then, the sequence S is calculated using formula (2) n1 , from the sequence S n1 Find the time corresponding to the k largest peaks, recorded as Step 2.4: Calculate C using formula (3) in, Representing a collection The number of elements in; replace X(n0) in formula (2) with C to reconstruct the final release sequence Step 2.5: Setup Step 2.6: Repeat steps 2.3-2.5 multiple times to ensure that as many release sequences as possible can be reconstructed, and classify and organize the obtained release sequences.

3. The transfer entropy coupling analysis method based on multi-channel surface electromyography signal decomposition according to claim 1 is characterized in that: In step 4, the signal is divided into alpha band, beta band, and gamma band.

4. The transfer entropy coupling analysis method based on multi-channel surface electromyography signal decomposition according to claim 1 is characterized in that: In step 5, the transfer entropy is calculated as follows: Two time series X = {x1, x2…, x n }, Y={y1,y2…,y n }, where n is the length of the time series, x i ,y i are the i-th observation values ​​respectively, then the transfer entropy from X to Y is expressed in biomedical signal analysis as p(·) represents the probability value; where τ is the time lag of X.

Images