Upper limb intermuscular coupling analysis method based on time-frequency-graph multi-domain high-order features

Abstract

The invention discloses an upper limb intermuscular coupling analysis method based on time-frequency-graph multi-domain high-order features. The method comprises the steps that S1, multi-channel sEMG signals are collected and preprocessed; s2, calculating time domain OIR under each order muscle combination under a natural operation normal form, including S2-1, for a dynamic process system V = {V1, V2... VN} composed of N channels of sEMG signals, the dynamic process of the sEMG signals of each channel is expressed as Vi = Vi1, Vi2... Vin}; the entropy rate of the dynamic processes and the joint entropy rate S2-2 of the two dynamic processes are calculated, and the time domain OIR of the dynamic process system V is obtained according to the organization structure information and the entropy rate of the intermuscular correlation network; s3, quantifying the contribution of the specific muscle in the multi-channel IMC system by calculating gradient OIR; s4, quantizing frequency domain distribution of high-order interaction of the multi-channel IMC system by calculating frequency OIR; and S5, calculating a redundancy cooperative balance index rate, and constructing an IMC high-order relation network according to the redundancy cooperative balance index rate.

Description

Technical Field

[0001] This invention relates to a method for analyzing intermuscular coupling under the natural operation paradigm of the upper limb, and particularly to a coupling analysis method based on multi-domain higher-order structural features. Background Technology

[0002] Human movement is a complex process regulated by the central nervous system (CNS). The CNS transmits motor control signals to relevant muscles via neural oscillations, resulting in oscillatory activity of motor units. Intermuscular coupling (IMC) refers to the information transmission and functional connectivity between muscles at multiple spatiotemporal levels, regulated by functional modulation and feedback control of the CNS. Surface electromyography (sEMG) signals contain action potentials of numerous motor units under the control of the CNS. IMC analysis based on sEMG not only reflects the interactions between muscles during movement but also quantifies the corticospinal drive shared between motor neurons, providing a theoretical basis for understanding the organization and coordination of motor control. In natural maneuvers, there are typically two motor components: a grasping component that activates the forearm muscles and an elevating component that activates the proximal arm muscles. Maintaining these tasks requires coordinated activation of the entire arm muscles, which involves complex and precise neural control of the muscles.

[0003] Current IMC analyses primarily rely on pairwise measurements to assess coupling by quantifying the information transfer or connection strength between pairs of sEMG channels. For example, methods based on coherence, Granger causality, and maximum information coefficients are used to identify changes in muscle connectivity. These methods calculate the coupling strength between paired sEMG signals, constructing muscle functional connectivity networks to reveal specific response patterns of the central nervous system to changes in grip strength. However, these methods fail to fundamentally capture the ubiquitous high-order interactions (HOIs) in neural connectivity, which involve collective behaviors arising from interactions between more than two nodes. Furthermore, they struggle to distinguish between direct connections between nodes and indirect connections resulting from shared drives or cascade effects, making it difficult to reveal the connectivity structure of intermuscular coupling networks.

[0004] To address the aforementioned issues, multivariate information theory offers several advanced measurement methods capable of capturing system characteristics beyond pairwise correlations. For example, the partial information decomposition (PID) framework decomposes mutual information into fundamental components that describe higher-order dependencies between three or more random variables: redundancy and cooperability. Redundant information refers to information shared among multiple variables, fully obtainable from any single variable, while cooperability information is specific to the entire multivariate system and cannot be obtained from a single variable. The emergence of organizational structure information resolves the computational complexity of the PID method. By measuring the difference between total correlation and dual total correlation in a multivariate system, it can characterize the balance of redundancy and cooperability in the system, and its computational efficiency is more suitable for large-scale systems. However, organizational structure information also has its limitations. It can only measure the overall redundancy / cooperability properties of higher-order systems, lacking a description of the network connectivity characteristics of higher-order interaction networks. Furthermore, the implementation of organizational structure information primarily focuses on static multivariate networks, potentially neglecting the dynamic characteristics of the system. Therefore, a novel method for analyzing higher-order interactions of inter-variable coupling is needed to measure the higher-order interaction characteristics of inter-variable coupling from multiple perspectives while preserving the dynamic characteristics of the system. Summary of the Invention

[0005] To address the shortcomings of existing technologies, such as neglecting the dynamic characteristics of information transmission and the frequency domain characteristics of higher-order structures, this invention proposes a time-frequency-graph multi-domain IMC higher-order structure analysis framework based on the tissue structure information rate and the redundancy synergistic balance index rate. This framework aims to capture the time-domain and frequency-domain characteristics of higher-order interactions and connection structures in intermuscular coupling.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for analyzing intermuscular coupling in the upper limb based on time-frequency-graph multi-domain high-order features includes the following steps:

[0008] S1. Collect and preprocess multi-channel sEMG signals under the natural operation paradigm of the upper limb to form a multi-channel IMC system;

[0009] S2. Calculate the time-domain OIR of muscle combinations of each order under the natural operating paradigm, including:

[0010] S2-1. For a dynamic process system V = {V1, V2, ..., V...} composed of N channels of sEMG signals, N}, where the dynamic process of the sEMG signal in each channel is represented by V. i ={V i1 V i2 …,V in}; Calculate the entropy rate of the dynamic process. and the joint entropy rate of the two dynamic processes

[0011] S2-2, Based on the tissue structure information of the intermuscular correlation network and the entropy rate... Obtain the time-domain OIR of the dynamic process system V;

[0012] S3. Quantify the contribution of a specific muscle in the multichannel IMC system by calculating the gradient OIR;

[0013] S4. Quantize the frequency domain distribution of the high-order interactions of the multi-channel IMC system by calculating the frequency OIR.

[0014] S5. Calculate the redundancy collaborative balance index rate and construct an IMC high-order relationship network based on the redundancy collaborative balance index rate;

[0015] Preferably, step S3 includes:

[0016] S3-1. Calculate the gradient tissue structure information between muscle nodes in the intermuscular correlation network;

[0017] S3-2, According to the entropy rate and the joint entropy rate Obtain the mutual information rate between the sEMG signals of two channels per unit time:

[0018]

[0019] Among them, entropy rate The calculation formula is:

[0020]

[0021] The joint entropy rate of the two sEMG signals is:

[0022]

[0023] S3-3. Generalize the gradient organization structure information based on the mutual information rate to calculate the gradient OIR:

[0024]

[0025] in For muscle V N With the remaining muscular system V N-1 Mutual information rate between them For muscle V N With removal of muscle V i Subsequent systems The mutual information rate between them.

[0026] Preferably, step S4 includes:

[0027] S4-1. Obtain the dynamic process system V = {V1, V2, ..., V...} N Any two muscles V in} i and V j Frequency domain mutual information rate between:

[0028]

[0029] Where ω∈[-Π, Π] is the normalized angular frequency. Represents the joint power spectral density matrix:

[0030]

[0031] in and Representing V respectively i and V j The power spectral density matrix, and is the cross-power spectral density matrix, and represents the conjugate relation;

[0032] S4-2. Obtain the frequency OIR based on the frequency domain mutual information rate:

[0033]

[0034] in The frequency domain gradient OIR is defined as:

[0035]

[0036] in For muscle V N With the remaining muscular system V N-1 Frequency mutual information rate between them For muscle V N With removal of muscle V i

[0037] Subsequent systems The frequency mutual information rate between them.

[0038] Preferably, step S5 includes:

[0039] S5-1, Obtaining the conditional mutual information rate:

[0040]

[0041] in, Represents the conditional mutual information rate between Vi and the system {Vj,Vk};

[0042] {V i V j} and V k The rate of interaction information between them is defined as:

[0043]

[0044] S5-2. Obtain the redundancy cooperative balance index rate based on the conditional mutual information rate:

[0045]

[0046] in, Corresponding to the extreme case of redundancy, i.e., V i and V j The interactions between them are entirely explained by other variables in the system; conversely, This represents the maximum synergistic effect, indicating that V i and V j The interactions between them fully reflect their effects on other variables in the system; This indicates that the redundancy and synergy between these two variables are balanced; if mean and All are 0, indicating that V i and V j There is no interaction between them.

[0047] Preferably, the method further includes step S6, separating the IMC high-order relationship network into a redundant network and a cooperative network.

[0048] Preferably, step S6 includes:

[0049] The connectivity of the IMC higher-order relation network is measured using weighted node degree, where each weighted node degree is the average of all edges connected to that node:

[0050]

[0051] Where N is the number of nodes, A ij It is the edge connecting node i and node j;

[0052] Using the node clustering coefficient WCC i Describe the degree of clustering among nodes in the network:

[0053]

[0054] As a preferred option, it also includes S7, classifying different upper limb movements.

[0055] Preferably, step S7 includes:

[0056] S7-1. Obtain the topological features of the redundant network and the cooperative network; the topological features include the weighted node degree, the node clustering coefficient, and modularity features; the Lasso algorithm is used to filter the topological features;

[0057] S7-2. Select Support Vector Machine as the classifier to classify different upper limb movements; use 5-fold cross-validation to verify the classification effect; evaluate the classification performance by calculating the classification accuracy and F1 score of the classification task.

[0058] Compared with the prior art, the beneficial effects of the present invention are reflected in:

[0059] 1. Unlike current intermuscular coupling analysis methods that mainly use pairwise metrics to quantify muscle connectivity, this invention proposes an upper limb intermuscular coupling analysis method based on higher-order structures, starting from multivariate information theory, which reveals the differences in higher-order interaction patterns in natural upper limb maneuvering tasks.

[0060] 2. Unlike existing high-order measurement methods that can only indirectly evaluate the high-order characteristics of a system through correlation networks, this invention uses entropy rate instead of Shannon entropy to extend organizational structure information to dynamic process systems. It can directly measure the high-order interactions between multi-channel sEMG signals, thereby preserving the potential dynamic characteristics of sEMG signals.

[0061] 3. In order to fully reflect the high-order interaction modes of intermuscular coupling systems, this invention quantifies the redundancy / cooperation modes of high-order muscle combinations from three perspectives: time domain, frequency domain, and network connectivity, providing a more novel and comprehensive perspective for intermuscular coupling analysis. Attached Figure Description

[0062] Figure 1 This is a flowchart of the method in Embodiment 1 of the present invention;

[0063] Figure 2 The time-domain OIR statistics of muscle combinations of various orders in Embodiment 1 of the present invention;

[0064] Figure 3 This contributes to the redundant information of each muscle in the intermuscular coupling system of Embodiment 1 of the present invention;

[0065] Figure 4 This is the frequency distribution of higher-order interactions of muscle combinations in Embodiment 1 of the present invention;

[0066] Figure 5 This is the BIR connection matrix of Embodiment 1 of the present invention;

[0067] Figure 6 This is the redundant / cooperative network structure of Embodiment 1 of the present invention. Detailed Implementation

[0068] To make the technical means, inventive features, objectives, and effects of the invention readily understandable, the invention is further described below with reference to specific illustrations. However, the invention is not limited to the embodiments described below.

[0069] It should be noted that the structures, proportions, sizes, etc., illustrated in the accompanying drawings of this specification are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0070] Example 1:

[0071] like Figure 1 The upper limb intermuscular coupling analysis method based on time-frequency-graph multi-domain high-order features, as shown, includes the following steps:

[0072] S1. Acquire multi-channel sEMG signals under the natural operation paradigm of the upper limb;

[0073] Six-channel surface electromyography (sEMG) signals were acquired using an electromyography device under the natural operation paradigm of the upper limb, including gripping (G), lifting (L), and gripping & lifting (G & L). The natural operation paradigm consists of three forearm muscles: flexor radialis (FCR), flexor superficialis (FDS), and flexor extensor lateralis (EDC) and three proximal muscles: biceps brachii (BB), triceps brachii (TB), and anterior deltoid (AD). The sampling frequency was 1000 Hz.

[0074] Preprocessing is performed on the multi-channel sEMG signal. A 4th-order zero-phase Butterworth filter is used to perform bandpass filtering on the multi-channel sEMG signal from 5 to 200 Hz, and a notch filter is used at 50 Hz to remove power line noise.

[0075] S2. Calculate the information rate about organizational structure (OIR) for different muscle combinations from level 1 to level 6, and compare the temporal characteristics of higher-order intermuscular interactions between different muscle combinations and the natural upper limb manipulation paradigm.

[0076] The range of muscle combinations is from 1 to 6, including single muscle combinations, combinations of two muscles, combinations of three muscles, and combinations of all six muscles. OIR (Organizational Interaction) is a measure of high-order interactions between muscles, reflecting their synergistic effects in different movements. It reveals the complex coupling relationships between muscles by analyzing the high-order statistical properties of muscle signals.

[0077] Extending organizational structure information to dynamic processes through entropy rate; including the following specific steps;

[0078] S2-1. Obtain the entropy rate of the dynamic process of the sEMG signal of each channel in the dynamic process system.

[0079] As a fundamental metric in information theory, Shannon entropy is used to quantify the information contained in a random variable X:

[0080]

[0081] Where p(*) represents the marginal distribution, This is the statistical expectation. Based on Shannon entropy, conditional entropy is defined as:

[0082] H(X i |X j )=H(X i ,X j )-H(X j ),

[0083] Where H(X) i ,X j ) is the variable X i and X j The joint entropy of the system is calculated using the following formula:

[0084]

[0085] Where p(x) i ,x j ) represents variable X i and X j The joint distribution probability.

[0086] Since Shannon entropy can only quantify the information carried by random variables, it cannot adapt to dynamic stochastic processes. Therefore, it potentially loses the dynamic characteristics of inter-variable coupling. To solve this problem, entropy rate is used to quantify the rate at which new information is generated in a dynamic process, replacing traditional Shannon entropy. Entropy rate is used to describe the average rate at which a stochastic process generates new information per unit time. For a dynamic process system V = {V1, V2, ..., V...} composed of N channels of sEMG signals... N}, where the dynamic process of the sEMG signal in each channel is represented by V. i ={V i1 Vi2 …,V in Its entropy rate is expressed as:

[0087]

[0088] Similarly, the joint entropy rate of the stochastic process variables of two dynamic processes is:

[0089]

[0090] S2-2, Obtain the time-domain OIR of a multi-channel sEMG system

[0091] Organizational structure information can be used to quantify redundancy or synergy in static network systems. A static network system refers to an intermuscular correlation network composed of multiple muscles. For example, for an intermuscular correlation network composed of N muscles, the organizational structure information (OI) of the network is:

[0092]

[0093] The superscript X N Represents a system consisting of N nodes, with the subscript X. i Represents a node with any variable, while Represents X N Remove node X i The same applies to subsets below. If Ω > 0, it indicates that the higher-order interactions of the static network system are redundant; conversely, if Ω < 0, the static network system mainly exhibits synergistic effects; if Ω = 0, the redundancy and synergy within the static network system are in balance.

[0094] Based on the entropy rate, OI is extended to dynamic process systems, and the time-domain OIR is defined as:

[0095]

[0096] Since second-order systems do not have higher-order interactions, their OIR is defined as 0, i.e. Similar to organizational structure information, a positive OIR value indicates redundancy, while a negative value indicates synergy. For example... Figure 2 As shown, numbers 1-6 represent FCR, FDS, EDC, BB, TB, and AD muscles, respectively. Under the upper limb natural maneuvering paradigm, higher-order muscle interactions are primarily dominated by redundant information, but differences exist in the higher-order dependency patterns between different muscle combinations. Both proximal and forearm muscle groups exhibit redundant higher-order dependencies, while synergistic effects mainly occur in combinations of proximal and forearm muscles. Significant differences exist in higher-order muscle interaction patterns across different tasks. G&L is a more functionally demanding motor task, thus requiring the integration of muscle activity across the entire arm, leading to higher redundancy in proximal and forearm muscle groups.

[0097] S3. Quantify the redundancy or synergistic contribution of specific muscles in a multi-channel IMC system using gradient OIR:

[0098] S3-1. Gradient tissue structure information can measure the properties of specific muscle nodes in intermuscular correlation networks, defined as:

[0099]

[0100] in Represents from set X N Remove {X} from the middle i ,X k The subset following}

[0101] When ΔΩ is positive, it indicates that the muscle introduces redundant information into the intermuscular correlation network. If ΔΩ is negative, the muscle has a cooperative relationship with the other muscles.

[0102] S3-2. Obtain the mutual information rate between two sEMG signals per unit time.

[0103] Static information theory, based on Shannon entropy, uses mutual information to measure the shared information between variables:

[0104]

[0105] To preserve the dynamic characteristics of the system and quantify the amount of information shared between two muscles per unit time, a mutual information rate is defined based on the entropy rate to represent the interaction between two sEMG signals.

[0106]

[0107] S3-3. To preserve the dynamic characteristics of the system, the mutual information rate is used to generalize the gradient organization structure information, and the gradient OIR is proposed, defined as:

[0108]

[0109] Similar to gradient tissue structure information, the positive or negative value of the gradient OIR also corresponds to the redundancy / co-contribution of the muscle.

[0110] Gradient OIR was used to quantify and compare the redundant or synergistic contributions of six different muscles in the intermuscular coupling system. Figure 3The gradient OIR of a 6th-order muscle system under different movement tasks is shown. In the G task, the FCR, FDS, and EDC muscles in the muscle coupling system contribute relatively high levels of redundant information. Conversely, in the L task, the BB, TB, and AD muscles contribute relatively high levels of redundant information. In the G&L tasks, the contributions of each muscle to the redundant information in the muscle coupling system are relatively balanced. This is because the G task mainly relies on forearm muscles to complete the movement, while the L task mainly relies on proximal muscles for force exertion and stability. The G&L tasks, however, require the coordinated action of multiple muscle joints.

[0111] S4. Calculate higher-order interactions in the 0-100Hz frequency range using frequency OIR and plot the frequency domain redundancy co-operation curve, where numbers 1-6 represent FCR, FDS, EDC, BB, TB, and AD, respectively. Figure 4 As shown, the redundancy of higher-order muscle combinations tends to peak in the beta band, while synergy is more common in the gamma band.

[0112] S4-1. For a muscle group V = {V1, V2, ..., V...} with N muscles, ... N The intermuscular coupling system composed of} allows for the coupling of any two muscles V i and V j The frequency domain mutual information rate between them can be defined as:

[0113]

[0114] Where ω∈[-Π, Π] is the normalized angular frequency. Represents the joint power spectral density matrix:

[0115]

[0116] in and Representing V respectively i and V j The power spectral density matrix, and is the cross-power spectral density matrix, and represents the conjugate relation;

[0117] S4-2. Subsequently, the frequency OIR is defined as:

[0118]

[0119] in The frequency domain gradient OIR is defined as:

[0120]

[0121] S5. Construct a higher-order intermuscular relationship network;

[0122] S5-1. A redundancy-cooperative balance index is proposed to quantify the redundancy / cooperative balance relationship between variable pairs in high-order systems. The connection structure of HOIs networks is constructed, and the conditional mutual information rate is first defined as follows:

[0123]

[0124] in, This represents the mutual information rate between muscle Vi and the muscle system {Vj,Vk}.

[0125] Combining step 5-1, the difference between mutual information rate and conditional mutual information rate highlights the balance between statistical concepts of redundancy and synergy within the system. i V j} and V k The rate of interaction information between them is defined as:

[0126]

[0127] S5-2: By standardizing the interaction information rate, the redundancy / synergy balance index rate (BIR) is defined as:

[0128]

[0129] in, Corresponding to the extreme case of redundancy, i.e., V i and V j The interactions between them are entirely explained by other variables in the system. Conversely, This represents the maximum synergistic effect, indicating that V i and V j The interactions between them fully reflect their effects on other variables in the system. This indicates that the redundancy and synergy between these two variables have reached a balance. If It means I Vi；Vj；Vk and I Vi；Vj All are 0, which indicates that V i and V j There is no interaction between them. However, these extreme cases are rarely observed in real-world systems. Typically, the BIR value ranges between (-1, 0) and (0, 1), representing cooperative and redundant states, respectively.

[0130] The BIR (Body Integrity Regulator) is used to measure the redundancy or synergistic balance of paired muscle connections in higher-order intermuscular interactions, such as... Figure 5 and Figure 6As shown, in the G&L and L tasks, the redundancy between the three proximal muscles is relatively high. In the G task, the redundancy between forearm muscles is slightly higher than that between proximal muscles, while synergy mainly occurs in the cross-combination of proximal / forearm muscles. The high redundancy within the same muscle group may be due to shared neural input, which ensures the robustness of gross tasks. The synergistic information between different muscles reflects multi-joint muscle coordination under the control of the central nervous system.

[0131] S6. Separate the IMC higher-order relationship network into redundant subnetworks and cooperative subnetworks. Use weighted node degree to measure the connectivity of the IMC higher-order relationship network, with values ​​ranging from [0,1]. The higher the node degree, the stronger the node's role in constructing and maintaining effective information. The degree of each node is the average of all edges connected to that node.

[0132]

[0133] Where N is the number of nodes, A ij It is the connecting edge between node i and node j.

[0134] The node clustering coefficient preserves the weight information of all connected edges and is used to describe the degree of clustering between nodes in the network. A higher clustering coefficient indicates better network connectivity for that node, providing a relative measure (Ai) of the connectivity between node i and its neighboring nodes. ij A ik It also considers the interconnectivity A of adjacent nodes. jk :

[0135]

[0136] S7. Classify different upper limb movements using higher-order connectivity features. The specific steps are as follows:

[0137] S7-1. Based on the topological features of redundant and cooperative sub-networks, node degree, clustering coefficient and modular features were extracted from 33 tasks of all subjects. Considering that high-dimensional features require a large amount of computation and may lead to overfitting, the Lasso algorithm was used to screen the features.

[0138] S7-2. Select Support Vector Machine as the classifier. Use 5-fold cross-validation to obtain more reliable classification results. To evaluate classification performance, calculate the classification accuracy and F1 score for each classification task.

[0139] Table 1 shows the classification results of different network topological features in the natural operation task. The results show that, compared with the cooperative network, the redundant network has a stronger ability to identify the difference between grasping and lifting tasks, with a classification accuracy of 81.82%. In the other two tasks, the cooperative network achieved accuracies of 90.91% and 86.36%, respectively.

[0140] Table 1: Upper limb movement binary classification results of different subnetworks

[0141]

[0142] Table 2 shows the three-class classification results for different networks. The results show that the redundancy subnetwork and the cooperative subnetwork achieved accuracies of 73.52% and 75.76% respectively in the three-class classification task, while their F1 scores were 71.14% and 75.58% respectively.

[0143] Table 2: Upper limb movement tri-classification results of different sub-networks

[0144]

Claims

1. A method for analyzing intermuscular coupling in the upper limb based on time-frequency-graph multi-domain high-order features, characterized in that, Includes the following steps: S1. Collect and preprocess multi-channel sEMG signals under the natural operation paradigm of the upper limb to form a multi-channel IMC system; S2. Calculate the time-domain OIR of muscle combinations of each order under the natural operating paradigm, including: S2-1. For a dynamic process system V = {V1, V2, ..., V...} composed of N channels of sEMG signals, N }, where the dynamic process of the sEMG signal in each channel is represented by V. i ={V i1 V i2 …,V in }; Calculate the entropy rate of the dynamic process. and the joint entropy rate of the two dynamic processes S2-2, Based on the tissue structure information of the intermuscular correlation network and the entropy rate... Obtain the time-domain OIR of the dynamic process system V; S3. Quantify the contribution of a specific muscle in the multichannel IMC system by calculating the gradient OIR; S4. Quantize the frequency domain distribution of the high-order interactions of the multi-channel IMC system by calculating the frequency OIR. S5. Calculate the redundancy collaborative balance index rate and construct the IMC high-order relationship network based on the redundancy collaborative balance index rate.

2. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 1, characterized in that, Step S3 includes: S3-1. Calculate the gradient tissue structure information between muscle nodes in the intermuscular correlation network; S3-2, According to the entropy rate and the joint entropy rate Obtain the mutual information rate between the sEMG signals of two channels per unit time: Among them, entropy rate The calculation formula is: The joint entropy rate of the two sEMG signals is: S3-3. Generalize the gradient organization structure information based on the mutual information rate to calculate the gradient OIR: in For muscle V N With the remaining muscular system V N-1 Mutual information rate between them For muscle V N With removal of muscle V i Subsequent systems The mutual information rate between them.

3. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 1, characterized in that, Step S4 includes: S4-1. Obtain the dynamic process system V = {V1, V2, ..., V...} N Any two muscles V in} i and V j Frequency domain mutual information rate between: Where ω∈[-Π, Π] is the normalized angular frequency. Represents the joint power spectral density matrix: in and Representing V respectively i and V j The power spectral density matrix, and is the cross-power spectral density matrix, and represents the conjugate relation; S4-2. Obtain the frequency OIR based on the frequency domain mutual information rate: in The frequency domain gradient OIR is defined as: in For muscle V N With the remaining muscular system V N-1 Frequency mutual information rate between them For muscle V N With removal of muscle V i Subsequent systems The frequency mutual information rate between them.

4. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 1, characterized in that, Step S5 includes: S5-1, Obtaining the conditional mutual information rate: I Vi；Vj|Vk =I Vi；Vj,Vk -I Vi；Vk , Among them, I Vi；Vj,Vk Represents the conditional mutual information rate between Vi and the system {Vj,Vk}; {V i V j } and V k The rate of interaction information between them is defined as: I Vi；Vj；Vk =I Vi；Vj -I Vi；Vj |V k . S5-2. Obtain the redundancy cooperative balance index rate based on the conditional mutual information rate: in, Corresponding to the extreme case of redundancy, i.e., V i and V j The interactions between them are entirely explained by other variables in the system; conversely, This represents the maximum synergistic effect, indicating that V i and V j The interactions between them fully reflect their effects on other variables in the system; This indicates that the redundancy and synergy between these two variables are balanced; if It means I Vi；Vj；Vk and I Vi；Vj All are 0, indicating that V i and V j There is no interaction between them.

5. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 1, characterized in that, It also includes S6, separating the IMC high-order relationship network into a redundant network and a cooperative network.

6. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 5, characterized in that, Step S6 includes: The connectivity of the IMC higher-order relation network is measured using weighted node degree, where each weighted node degree is the average of all edges connected to that node: Where N is the number of nodes, A ij It is the edge connecting node i and node j; Using the node clustering coefficient WCC i Describe the degree of clustering among nodes in the network: Among them, A ij It is the connecting edge between node i and node j; A ik A is the edge connecting node i and node k; jk It is the connecting edge between node k and node j.

7. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 6, characterized in that, It also includes S7, which categorizes different upper limb movements.

8. The method for upper limb intermuscular coupling analysis based on time-frequency-graph multi-domain high-order features according to claim 7, characterized in that, Step S7 includes: S7-1. Obtain the topological features of the redundant network and the cooperative network; the topological features include the weighted node degree, the node clustering coefficient, and modularity features; the Lasso algorithm is used to filter the topological features; S7-2. Select Support Vector Machine as the classifier to classify different upper limb movements; use 5-fold cross-validation to verify the classification effect; evaluate the classification performance by calculating the classification accuracy and F1 score of the classification task.

Images