Accelerated rehabilitation data processing method and device based on electrophysiological data

Abstract

The invention discloses an accelerated rehabilitation data processing method and device based on electrophysiological data, and relates to the technical field of medical rehabilitation, and the method comprises the steps: calculating a full-mode phase locking value based on a noise reduction matrix, and generating a fusion feature vector through a gating circulation network; microcosmic entropy, mesoscopic entropy, macroscopic entropy and psychological entropy are extracted from the fusion feature vector, and an entropy state vector is constructed; according to the entropy state vector, generating personalized rehabilitation parameters under security constraints and exploration constraints; the personalized rehabilitation parameters are converted into exoskeleton control instructions and virtual reality scene parameters, and feedback data are collected and executed; and dynamically updating parameters of signal noise reduction, feature fusion and strategy optimization based on execution feedback data. According to the method, the quality and precision of multi-mode electrophysiological signal processing are improved, optimization of personalized rehabilitation parameters is achieved, the key problems in the prior art are solved through the innovations, and wide application prospects are shown in the cross field of biomedical engineering and artificial intelligence.

Description

Technical Field

[0001] The present invention relates to the technical field of medical rehabilitation, and in particular to an accelerated rehabilitation data processing method and device based on electrophysiological data. Background Art

[0002] In recent years, with the development of biomedical engineering technology, the application of electrophysiological data in rehabilitation treatment has become increasingly widespread. The acquisition and analysis of multi-modal data such as surface electromyogram signal (sEMG), electrocardiogram signal (ECG), electro-dermal activity signal (EDA), and acceleration signal have become research hotspots. These technologies can provide rich information for evaluating and guiding the rehabilitation process by monitoring the physiological states of different parts of the human body. However, traditional electrophysiological data analysis methods mainly rely on a single type of signal and lack effective noise reduction and feature fusion strategies, resulting in incomplete and inaccurate information obtained, which is difficult to meet the needs of personalized rehabilitation. In addition, existing technologies often ignore the dynamic correlation between signals when processing cross-modal signals, restricting the ability to understand and respond to complex rehabilitation needs.

[0003] Specifically, there are two main deficiencies in the analysis systems based on single-modal or multi-modal electrophysiological signals currently used in the rehabilitation field. First, traditional methods usually adopt fixed filters or predefined noise reduction algorithms, which are inadequate in the face of complex actual application scenarios because the noise patterns vary significantly under different individuals and environmental conditions. Second, in the feature extraction process, most existing technologies focus on single-dimensional data analysis and fail to fully utilize the interaction between multiple signals to reveal deep physiological state changes. In contrast, an accelerated rehabilitation data processing method and system based on electrophysiological data proposed by the present invention not only uses an adaptive noise reduction algorithm to generate a time-aligned noise reduction matrix, but also calculates the cross-modal phase locking value and uses a gated recurrent network to generate a fused feature vector, effectively solving the above problems. This method can more accurately capture the true physiological state of patients and provide a scientific basis for formulating personalized rehabilitation plans. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an accelerated rehabilitation data processing method based on electrophysiological data to solve the problems in the prior art of being unable to effectively integrate multi-sensor data and lacking a mechanism for dynamically adjusting personalized rehabilitation parameters.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an accelerated rehabilitation data processing method based on electrophysiological data, which includes synchronously collecting surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals, and generating a time-aligned noise reduction matrix through an adaptive noise reduction algorithm; calculating the full-modal phase locking value based on the noise reduction matrix, and generating a fused feature vector through a gated recurrent network; extracting micro entropy, meso entropy, macro entropy, and psychological entropy from the fused feature vector, and constructing an entropy state vector; generating personalized rehabilitation parameters under safety constraints and exploration constraints according to the entropy state vector; converting the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters, and collecting execution feedback data; dynamically updating the parameters of signal noise reduction, feature fusion, and strategy optimization based on the execution feedback data.

[0008] As a preferred embodiment of the accelerated rehabilitation data processing method based on electrophysiological data of the present invention, the specific steps of generating a time-aligned noise reduction matrix through the adaptive noise reduction algorithm are as follows:

[0009] Perform noise reduction processing on the collected surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals;

[0010] Align the time and generate a noise reduction matrix through hardware-level synchronization, resampling alignment, matrix construction, and data verification.

[0011] As a preferred embodiment of the accelerated rehabilitation data processing method based on electrophysiological data of the present invention, the specific steps of calculating the full-modal phase locking value based on the noise reduction matrix and generating a fused feature vector through the gated recurrent network are as follows:

[0012] Extract the time-frequency phase of the surface electromyogram signal, the piecewise linear phase of the electrocardiogram signal, the exponential decay phase of the skin electrical activity signal, and the phase-locked loop tracking phase of the acceleration signal from the noise reduction matrix, and generate a phase matrix by aligning them along the time axis;

[0013] Calculate the full-modal phase locking value based on the phase matrix through a multi-modal PLV formula;

[0014] Define a node set and a hyperedge set based on the full-modal phase locking value, and construct a dynamic hypergraph through a hypergraph adjacency matrix;

[0015] Allocate the input weights of the signal channels based on the input gate and the forget gate according to the full-modal phase locking value, and adjust the forgetting ratio of the historical hidden state according to the acceleration phase difference;

[0016] Combine the input gate weight and the historical hidden state adjusted by the forget gate to generate a candidate hidden state and update the hidden state;

[0017] Aggregate the hidden states through hypergraph-constrained multi-head attention to generate a fused feature vector.

[0018] As a preferred solution of the accelerated rehabilitation data processing method based on electrophysiological data according to the present invention, wherein: extracting micro entropy, mesoscopic entropy, macroscopic entropy and psychological entropy from the fusion feature vector, and constructing an entropy state vector, the specific steps are as follows,

[0019] Extract entropy features from the fusion feature vector from four levels of micro, mesoscopic, macroscopic and psychological to generate micro entropy, mesoscopic entropy, macroscopic entropy and psychological entropy;

[0020] Construct an entropy state vector through the extracted micro entropy, mesoscopic entropy, macroscopic entropy and psychological entropy.

[0021] As a preferred solution of the accelerated rehabilitation data processing method based on electrophysiological data according to the present invention, wherein: generating personalized rehabilitation parameters according to the entropy state vector under safety constraints and exploration constraints, the specific steps are as follows,

[0022] Set safety thresholds C1, C3 and C4 and a time decay factor based on the historical rehabilitation data set;

[0023] Dynamically limit high-risk actions through safety constraint modeling, including,

[0024] When the micro entropy exceeds the safety threshold C1, reduce the training intensity parameter through the Sigmoid decay function;

[0025] When the macroscopic entropy exceeds the safety threshold C3, limit the joint range of motion parameter through the exponential truncation function;

[0026] When the psychological entropy exceeds the safety threshold C4, reduce the movement speed parameter through the piecewise linear reset function;

[0027] Encourage diverse action patterns through exploration constraint modeling, including,

[0028] Generate a training intensity exploration weight through the hyperbolic tangent function, generate a joint range of motion exploration weight through the inverse exponential function, generate a movement speed exploration weight through the power function, and superimpose the time decay factor;

[0029] Element-wise multiply the training intensity parameter by the training intensity exploration weight to generate a preliminary training intensity parameter;

[0030] Element-wise multiply the joint range of motion safety parameter by the joint range of motion safety exploration weight to generate a preliminary joint range of motion parameter;

[0031] Element-wise multiply the movement speed safety parameter by the movement speed safety exploration weight to generate a preliminary movement speed parameter;

[0032] Superimpose the exploration weights of training intensity, joint range of motion, and movement speed on the corresponding parameters respectively to form enhanced, adaptive, and driving intermediate parameters, and then perform a secondary superposition to generate the final personalized rehabilitation parameters.

[0033] As a preferred embodiment of the accelerated rehabilitation data processing method based on electrophysiological data according to the present invention, wherein: the step of converting the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters and collecting execution feedback data is as follows.

[0034] Convert the joint range of motion parameters into a spatial coordinate sequence through a cubic spline interpolation algorithm;

[0035] Convert the training intensity parameters into the motor torque output threshold through a fuzzy PID controller;

[0036] Convert the movement speed parameters into the joint angular velocity constraint value through the inverse solution algorithm of the Jacobian matrix;

[0037] Use the spatial coordinate sequence, the motor torque output threshold, and the joint angular velocity constraint value as the exoskeleton control instructions;

[0038] Convert the joint range of motion parameters into the movement amplitude of the virtual limb model through a non-linear amplification mapping rule;

[0039] Convert the movement speed parameters into the virtual obstacle generation frequency through a velocity-density correlation function;

[0040] Convert the training intensity parameters into a multi-modal prompt strategy through an intensity classification logic, and use the movement amplitude of the virtual limb model, the virtual obstacle generation frequency, and the multi-modal prompt strategy as the virtual reality scene parameters;

[0041] Perform exoskeleton drive execution and virtual scene rendering through the exoskeleton control instructions and the virtual reality scene parameters;

[0042] Synchronously collect the actual joint torque, trunk attitude angle, and fixation point coordinate data based on a six-dimensional force sensor array, an inertial measurement unit, and an eye movement tracking module.

[0043] As a preferred embodiment of the accelerated rehabilitation data processing method based on electrophysiological data according to the present invention, wherein: the step of updating the parameters of signal denoising, feature fusion, and strategy optimization based on the execution feedback data is as follows.

[0044] Generate a torque error scalar through the root mean square error between the actual joint torque and the expected torque of the exoskeleton, calculate the posture coordination error based on the dynamic time warping distance of the trunk attitude angle, and generate a visual attention error in combination with the virtual task fixation point overlap rate;

[0045] Dynamically adjust the wavelet packet sub-band noise threshold of the surface electromyogram signal according to the torque error scalar, correct the R-wave detection noise threshold of the electrocardiogram signal by median filtering based on the posture coordination error, and update the parameters of signal denoising, feature fusion and strategy optimization in combination with the visual attention error.

[0046] In a second aspect, the present invention provides an accelerated rehabilitation data processing device based on electrophysiological data, including an acquisition module for synchronously acquiring surface electromyogram signals, electrocardiogram signals, skin electrical activity signals and acceleration signals, and generating a time-aligned denoising matrix through an adaptive denoising algorithm; a fusion module for calculating the full-modal phase locking value based on the denoising matrix and generating a fusion feature vector through a gated recurrent network; a multi-scale entropy state analysis module for extracting micro entropy, meso entropy, macro entropy and mental entropy from the fusion feature vector and constructing an entropy state vector; a personalized rehabilitation strategy generation module for generating personalized rehabilitation parameters according to the entropy state vector under safety constraints and exploration constraints; an exoskeleton-VR execution control module for converting the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters and collecting execution feedback data; a dynamic parameter optimization module for dynamically updating the parameters of signal denoising, feature fusion and strategy optimization based on the execution feedback data.

[0047] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the accelerated rehabilitation data processing method based on electrophysiological data as described in the first aspect of the present invention is implemented.

[0048] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the accelerated rehabilitation data processing method based on electrophysiological data as described in the first aspect of the present invention is implemented.

[0049] The beneficial effects of the present invention are as follows: By synchronously acquiring and adaptively denoising, and calculating the full-modal phase locking value based on the denoising matrix and generating a fusion feature vector, the quality and accuracy of multi-modal electrophysiological signal processing are significantly improved. The former ensures the consistency and synchronization of the signals through precise time alignment and high-quality denoising; the latter realizes the optimization of personalized rehabilitation parameters through deep feature extraction and dynamic adjustment mechanisms. These innovations not only solve the key problems in the prior art, but also show broad application prospects in the cross-field of biomedical engineering and artificial intelligence. Description of the Drawings

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0051] Figure 1 It is a flowchart of the accelerated rehabilitation data processing method based on electrophysiological data in Embodiment 1;

[0052] Figure 2 It is a module diagram of the accelerated rehabilitation data processing device based on electrophysiological data in Embodiment 1. Specific Embodiments

[0053] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will provide a detailed description of the specific embodiments of the present invention with reference to the accompanying drawings of the specification.

[0054] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0055] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that can be included in at least one implementation manner of the present invention. The phrase "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments.

[0056] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides an accelerated rehabilitation data processing method based on electrophysiological data, including the following steps:

[0057] S1: Synchronously collect surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals, and generate a time-aligned noise reduction matrix through an adaptive noise reduction algorithm;

[0058] Furthermore, collect the electrophysiological signals of the target muscle group through an 8-channel flexible electrode array, synchronously collect the electrocardiogram signals of leads I, II, and III using a 3-lead Ag / AgCl electrode, collect the conductance change of the thenar muscle of the palm through a constant voltage (0.5V) sensor, and collect the limb movement acceleration through a 3-axis MEMS sensor;

[0059] Perform noise reduction processing on the collected surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals;

[0060] Denoise the surface electromyogram signal through an improved wavelet packet decomposition algorithm;

[0061] It should be noted that power interference is eliminated by a 50Hz notch filter. Each channel of surface electromyogram signal is decomposed into 8 sub - frequency bands through wavelet packet transform. The noise threshold is dynamically calculated for each sub - frequency band, soft - threshold processing is performed on each sub - frequency band, and the filtered sub - frequency bands are reconstructed into a denoised surface electromyogram signal;

[0062] Denoise the electrocardiogram signal through R - wave - driven adaptive comb filtering;

[0063] It should be noted that the R - wave in the QRS complex is located in real - time through the Pan - Tompkins algorithm. The specific steps are as follows: the QRS complex is enhanced through band - pass filtering (5 - 15Hz), the R - wave characteristics are highlighted using differentiation and squaring operations, a smooth envelope is generated through sliding window integration (150ms), and the R - wave position is determined through an adaptive threshold;

[0064] The adaptive threshold is divided into a detection threshold and a noise threshold. The detection threshold is used to determine valid R - waves, with an initial value of 50% of the average of the signal peaks in the first 2 seconds. The noise threshold is used to identify noise interference, with an initial value of 25% of the average of the signal peaks in the first 2 seconds;

[0065] When a valid R - wave is detected, the new detection threshold inherits 75% of the weight of the historical detection threshold and fuses 25% of the weight of the current peak, smoothly tracking the change in signal amplitude, and the noise threshold is updated synchronously;

[0066] If there is no valid R - wave continuously, the detection threshold gradually approaches the noise threshold to reduce the risk of missed detection, and the noise threshold is updated synchronously;

[0067] Design a filter according to the current RR interval (interval between adjacent R - waves) for denoising;

[0068] Denoise the skin electrical activity signal through acceleration - assisted motion artifact repair;

[0069] It should be noted that baseline drift correction is performed by eliminating low - frequency drift through 5 - second sliding window mean filtering, and motion artifact detection and repair are performed through artifact marking and linear interpolation repair;

[0070] Denoise the acceleration signal through low - pass filtering and gravity compensation;

[0071] It should be noted that high - frequency noise is suppressed by eliminating vibration noise through a 20Hz fourth - order Butterworth low - pass filter, and gravity compensation is performed by separating the static gravity component through quaternion attitude solution;

[0072] Align the time and generate a noise reduction matrix through hardware-level synchronization, resampling alignment, matrix construction, and data verification.

[0073] It should be noted that hardware-level synchronization is to achieve multi-device clock synchronization based on the IEEE 1588 PTP protocol, with a maximum deviation of <1ms. Resampling alignment is to perform cubic spline interpolation on surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals (EDA, ECG, acceleration) that are not sampled at 1000Hz and unify them to the 1000Hz time axis. Matrix construction is to merge the denoised surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals into a standardized matrix in chronological order. Data verification is to verify the signal alignment accuracy through cross-correlation analysis to ensure that the maximum time delay error is <5ms.

[0074] S2: Calculate the full-modal phase-locking value based on the noise reduction matrix and generate a fused feature vector through a gated recurrent network;

[0075] Furthermore, extract the time-frequency phase of the surface electromyogram signal, the piecewise linear phase of the electrocardiogram signal, the exponential decay phase of the skin electrical activity signal, and the phase-locked loop tracking phase of the acceleration signal from the noise reduction matrix, and generate a phase matrix by aligning them along the time axis;

[0076] Based on the phase matrix, calculate the full-modal phase-locking value through the multi-modal PLV formula. The expression is:

[0077]

[0078] Among them, PLV k,m,p,q (t) is the full-modal phase-locking value of the EMG channel k, ECG lead m, EDA signal, and acceleration axis q at time t. L is the sliding window length, τ is the time index variable, j is the imaginary unit, is the time-frequency phase of the k-th EMG channel at time point τ, is the piecewise linear phase of the m-th ECG lead at time point τ, φ EDA (τ) is the exponential decay phase of the EDA signal at time point τ, is the phase-locked loop tracking phase of the q-th acceleration axis at time point τ, and g(τ;t) is the spatio-temporal joint weight function;

[0079] It should be noted that, γ is the exponential decay factor, is the cosine smoothing term;

[0080] Define the node set and hyperedge set based on the full-modal phase-locking value and construct a dynamic hypergraph through the hypergraph adjacency matrix;

[0081] It should be noted that the node set is defined by regarding each signal channel as a hypergraph node (a total of 15 nodes: 8 EMG + 3 ECG + 1 EDA + 3 ACC) and forming a node set, and the hyperedge set is defined such that each hyperedge corresponds to a PLV quadruple (k, m, p, q) and forms a hyperedge set;

[0082] Based on the input gate and the forget gate, the input weights of the signal channels are assigned through the all-modal phase locking value, and the forgetting ratio of the historical hidden state is adjusted according to the acceleration phase difference;

[0083] It should be noted that the all-modal phase locking value tensor regulates the input gate weight assignment through the weighted summation process of the hypergraph weight matrix. Specifically, it is manifested as: the phase synchronization quadruples of the surface electromyogram channel, electrocardiogram lead, skin electrical activity signal, and acceleration axis are multiplied by the corresponding hypergraph weight matrix and then accumulated, and then superimposed on the linear transformation result of the input signal. After passing through the Sigmoid function, an input gate weight vector is generated. High-synchronization signal combinations (such as PLV > 0.8) obtain priority fusion weights; the acceleration phase difference is mapped to the forget gate bias term through linear transformation, and after being superimposed on the linear transformation result of the input signal, a forget gate weight vector is generated through the Sigmoid function. When the motion phase mismatch (such as the phase difference > π / 2), the forget gate weight approaches zero, and the non-coordinated motion noise in the historical hidden state is forcibly cleared;

[0084] Combined with the historical hidden state adjusted by the input gate weight and the forget gate, a candidate hidden state is generated and the hidden state is updated;

[0085] It should be noted that the input gate weight vector is multiplied element-wise with the current noise-reduced signal to screen out high-synchronization components. After linear transformation, it is added to the linear transformation result of the historical hidden state adjusted by the forget gate (the result of element-wise multiplication of the forget gate weight vector and the historical hidden state). A normalized candidate hidden state is generated through the hyperbolic tangent function; during the hidden state update process, the input gate weight vector controls the mixing ratio of the candidate hidden state and the historical hidden state. High-weight regions (such as input gate weight > 0.9) preferentially fuse the current signal, and low-weight regions (such as input gate weight < 0.1) retain historical information, and finally an updated hidden state vector is generated;

[0086] The hidden state is aggregated through the hypergraph-constrained multi-head attention to generate a fused feature vector.

[0087] It should be noted that the hypergraph adjacency matrix constructs a binary mask matrix to mark the effective interaction relationships of the internal nodes of the hyperedge (such as EMG channel 1 - ECG lead I - EDA - ACC X - axis); after the hidden state vector generates the query, key, and value vectors of each attention head through linear transformation, the attention weights of invalid node pairs are set to zero; the effective attention weights are calculated by scaling the query - key dot product, and after being filtered by the mask and passing through the softmax function to generate the normalized weights, the value vectors are weighted and summed and then concatenated with the multi - head output, and finally mapped to the fused feature vector through a linear projection matrix, only aggregating the feature information of physiological - related signal combinations (such as muscle activation and heart rate synchronization).

[0088] S3: Extract the micro - entropy, meso - entropy, macro - entropy, and psycho - entropy from the fused feature vector, and construct an entropy state vector;

[0089] Furthermore, based on the entropy features extracted from the fused feature vector at the micro, meso, macro, and psycho levels, the micro - entropy, meso - entropy, macro - entropy, and psycho - entropy are generated;

[0090] The micro - entropy is extracted through sub - band segmentation, complex Morlet wavelet transform, and permutation entropy calculation. The expression is:

[0091]

[0092] where \(S_1\) is the micro - entropy, is the \(u\) - th sub - band of the fused feature vector, is the complex Morlet wavelet transform, \(PE\) is the permutation entropy, \(\log_2(3!)\) is the most probable value of the permutation entropy, \(u\) is the sub - band index, \(f\) u is the central frequency, \(\sigma\) u is the bandwidth factor, \(o\) is the embedding dimension, and \(\delta\) is the delay parameter;

[0093] It should be noted that sub - band segmentation divides the fused feature vector into 4 sub - bands (20 - 100 Hz, 100 - 200 Hz, 200 - 300 Hz, 300 - 500 Hz). The complex Morlet wavelet transform performs time - frequency decomposition on each sub - band signal to extract the instantaneous energy distribution, and the permutation entropy calculation calculates the permutation entropy of each sub - band to measure the signal complexity;

[0094] The meso - entropy is extracted through Hurst exponent calculation, phase synchronization modulation, and detrended fluctuation analysis. The expression is:

[0095]

[0096] where \(S_2\) is the meso - entropy, is the Hurst exponent based on the RR interval, \(PS\) EMG-ECGis the electromyogram-electrocardiogram phase synchronization index, λ = 5 is the phase synchronization sensitivity adjustment parameter, A α is the detrended fluctuation analysis scaling exponent, α max = 1.2 is the maximum physiological scaling exponent;

[0097] It should be noted that the Hurst exponent is calculated based on the RR interval to calculate the long-term memory of the heart rate signal, the phase synchronization modulation is to dynamically adjust the Hurst exponent weight using the electromyogram-electrocardiogram phase synchronization index, and the detrended fluctuation analysis is to calculate the scaling exponent of a 500 ms window;

[0098] The macroscopic entropy is extracted through multi-scale entropy calculation and acceleration gradient detection, and the expression is:

[0099]

[0100] where S3 is the macroscopic entropy, is the acceleration signal of the 5-second window on the d-axis, MSE is the multi-scale entropy, s = 5 is the scale factor, r = 0.15 is the tolerance threshold, J d is the acceleration gradient, ∥·∥2 is the L2 norm;

[0101] It should be noted that the multi-scale entropy calculation is to perform multi-scale complexity analysis on the acceleration signal, and the acceleration gradient detection is to calculate the acceleration mutation detection index;

[0102] The psychological entropy is extracted through SCR event counting and phase synchronization analysis, and the expression is:

[0103]

[0104] where S4 is the psychological entropy, is the number of skin conductance response (SCR) events within the past 10 seconds, is the exponentially decaying phase of the EDA signal, is the R-wave triggered phase of the ECG signal;

[0105] It should be noted that the SCR event counting is to count the number of skin conductance responses within a 10-second window, and the phase synchronization analysis is to calculate the phase difference between EDA and ECG;

[0106] The entropy state vector is constructed by extracting the microscopic entropy, mesoscopic entropy, macroscopic entropy and psychological entropy.

[0107] It should be noted that based on the fused feature vector, the microscopic entropy is extracted by complex Morlet wavelet frequency division and permutation entropy to quantify the chaos degree of high-frequency electromyography signals within a 50-ms window; the mesoscopic entropy is extracted by the Hurst exponent and the dynamic weight of electromyogram-electrocardiogram phase synchronization to evaluate the autonomic sympathetic-vagal balance; the macroscopic entropy is extracted by multi-scale entropy and acceleration signal gradient analysis to quantify the movement coordination and joint mutation characteristics within a 5-s window; the psychological entropy is extracted by the fusion of the event density of skin conductance response (SCR) and the power of the phase difference between skin electrical activity and electrocardiogram to characterize the psychological stress level; finally, they are sequentially combined into a four-dimensional entropy state vector S1, S2, S3, S4, comprehensively covering the multi-scale physiological states of muscle activation, autonomic nerve regulation, movement control, and psychological load.

[0108] S4: Generate personalized rehabilitation parameters under safety constraints and exploration constraints according to the entropy state vector;

[0109] Set safety thresholds C1, C3, and C4 and the time decay factor based on the historical rehabilitation dataset;

[0110] It should be noted that by analyzing the correlation between multi-modal physiological signals (surface electromyogram, electrocardiogram, skin electrical activity, acceleration) and clinical safety events (muscle strain, joint compensation, anxiety trigger) in the historical rehabilitation dataset, the safety thresholds are determined:

[0111] The microscopic entropy safety threshold C1 = 0.7: When the microscopic entropy (muscle chaos degree) exceeds 0.7, the risk probability of muscle spasm or fatigue ≥ 75% (results of logistic regression analysis);

[0112] The macroscopic entropy safety threshold C3 = 0.6: When the macroscopic entropy (movement coordination) exceeds 0.6, the incidence rate of joint compensation actions increases significantly (p < 0.01, based on the statistics of motion capture data);

[0113] The psychological entropy safety threshold C4 = 0.8: When the psychological entropy (psychological stress level) exceeds 0.8, the probability of moderate or above anxiety in patients ≥ 90% (correlation verification with the anxiety scale GAD-7);

[0114] The time decay factor z = 0.95: Determined by optimizing through reinforcement learning, controlling the exploration intensity to decay by 5% every 10 minutes to balance the exploration and stability requirements;

[0115] Dynamically restrict high-risk actions through safety constraint modeling, including

[0116] When the microscopic entropy exceeds the safety threshold C1, reduce the training intensity parameter through the Sigmoid decay function;

[0117] When the macroscopic entropy exceeds the safety threshold C3, limit the joint activity range parameter through the exponential truncation function;

[0118] When the mental entropy exceeds the safety threshold C4, the movement speed parameter is reduced by a piecewise linear reset function;

[0119] It should be noted that when the microscopic entropy (muscle chaos) exceeds the safety threshold C1 = 0.7, the training intensity parameter is dynamically attenuated from the current value through the Sigmoid decay function. Specifically, the input of the Sigmoid function is the difference between the microscopic entropy and the safety threshold (S1 - C1), and the slope parameter is set to 10, so that the output value of the function rapidly decays from 1 to nearly 0 when S1 > C1. For example, when S1 = 0.8, the training intensity drops to about 30%, and when S1 = 0.9, it further drops to 5%, thus avoiding muscle spasms or fatigue injuries;

[0120] When the macroscopic entropy (S3, movement coordination) exceeds the safety threshold C3 = 0.6, the joint movement range parameter is restricted by an exponential truncation function. Specifically, the function form is When S3 = 0.7, the joint movement range is reduced to about 37%, and when S3 = 0.8, it is reduced to 13.5%, suppressing the risk of movement injuries caused by joint compensation or imbalance;

[0121] When the mental entropy (S4, mental stress level) exceeds the safety threshold C4 = 0.8, the movement speed parameter is reduced by a piecewise linear reset function. Specifically, in the interval 0.8 ≤ S4 < 0.9, the speed parameter decreases linearly from 1.0 to 0.5 according to the slope of 5; when S4 ≥ 0.9, the speed parameter is forced to zero and an emergency pause is triggered to prevent movement loss of control caused by anxiety or pain;

[0122] Encouraging diverse movement patterns through exploration constraint modeling, including,

[0123] Generating a training intensity exploration weight through the hyperbolic tangent function, generating a joint movement range exploration weight through the inverse exponential function, generating a movement speed exploration weight through the power function, and superimposing a time decay factor;

[0124] Element-wise multiplying the training intensity parameter by the training intensity exploration weight to generate a preliminary training intensity parameter;

[0125] Element-wise multiplying the joint movement range safety parameter by the joint movement range safety exploration weight to generate a preliminary joint movement range parameter;

[0126] Element-wise multiplying the movement speed safety parameter by the movement speed safety exploration weight to generate a preliminary movement speed parameter;

[0127] It should be noted that based on the mesoscopic entropy (S2, autonomic nerve balance), through the hyperbolic tangent function Generate the exploration weights for training intensity. When the vagus nerve is active (S2 approaches 1), the weight is close to 1.0. When the sympathetic nerve is active (S2 approaches 0), the weight is close to 0. The time decay factor z = 0.95 is superimposed to make the weight decay by 5% every 10 minutes;

[0128] Based on the macroscopic entropy (S3, motor coordination), through the inverse exponential function Generate the exploration weights for the range of joint motion. When the coordination is good (S3 approaches 0), the weight is close to 1.0. When the coordination is poor (S3 approaches 1), the weight approaches 0. The time decay factor z = 0.95 is superimposed to gradually reduce the exploration intensity;

[0129] Based on the psychological entropy (S4, psychological stress level), through the power function (1 - S4) 2 Generate the exploration weights for movement speed. When the psychological stress is low (S4 approaches 0), the weight is close to 1.0. When the stress is high (S4 approaches 1), the weight approaches 0. The time decay factor z = 0.95 is superimposed to suppress the long-term exploration risk;

[0130] Superimpose the exploration weights of training intensity, range of joint motion, and movement speed with the corresponding parameters respectively to form enhanced, adaptive, and driving intermediate parameters, and then perform a secondary superposition to generate the final personalized rehabilitation parameters.

[0131] It should be noted that the preliminary training intensity parameter is added to 30% of the training intensity exploration weight to generate the final training intensity parameter; the preliminary range of joint motion parameter is added to 30% of the exploration weight of the range of joint motion to generate the final range of joint motion parameter; the preliminary movement speed parameter is added to 30% of the exploration weight of movement speed to generate the final movement speed parameter. For example, if the training intensity exploration weight is 0.7, the independent superposition part is 0.7 × 30% = 0.21. When the preliminary training intensity parameter is 0.3, the final parameter is 0.3 + 0.21 = 0.51.

[0132] S5: Convert the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters, and collect the execution feedback data;

[0133] Furthermore, convert the range of joint motion parameter into a spatial coordinate sequence through the cubic spline interpolation algorithm;

[0134] Convert the training intensity parameter into the motor torque output threshold through the fuzzy PID controller;

[0135] Convert the movement speed parameter into the joint angular velocity constraint value through the inverse solution algorithm of the Jacobian matrix;

[0136] Use the spatial coordinate sequence, motor torque output threshold, and joint angular velocity constraint value as the exoskeleton control instructions;

[0137] It should be noted that interpolation nodes are defined based on the joint range of motion parameters, and a continuously differentiable joint angle - spatial coordinate mapping function is constructed using the cubic spline interpolation algorithm. A C 2 continuity constraint condition is imposed between adjacent nodes to generate a smooth trajectory spatial coordinate sequence; a fuzzy PID controller is designed for the training intensity parameter. The intensity error and its rate of change are quantified through membership functions, and a dynamic torque output threshold is generated by combining the fuzzy self - tuning rules of the proportional - integral - differential parameters. A saturation function is used to perform safety limiting on the output torque; a kinematic Jacobian matrix is constructed based on the exoskeleton link parameters, the generalized inverse of the Jacobian matrix is solved for the motion speed parameter, and the damping least - squares method is combined to handle the singular configuration. The Cartesian space velocity is mapped to the joint angular velocity constraint value, and a velocity threshold comparator is used to limit the angular velocity fluctuation range within ±3 rad / s.

[0138] The joint range of motion parameters are converted into the motion amplitude of the virtual limb model through a non - linear amplification mapping rule;

[0139] The motion speed parameters are converted into the virtual obstacle generation frequency through a velocity - density correlation function;

[0140] The training intensity parameters are converted into a multi - modal prompting strategy through an intensity classification logic;

[0141] The motion amplitude of the virtual limb model, the virtual obstacle generation frequency, and the multi - modal prompting strategy are used as virtual reality scene parameters;

[0142] It should be noted that a non - linear amplification mapping rule is defined based on the joint range of motion parameters, and a piece - wise exponential - linear composite function is used to handle the mapping relationship between the joint range of motion and the motion amplitude of the virtual limb model. When the joint range of motion parameter exceeds the preset joint angular velocity critical value, the exponential amplification mode is activated. When it is lower than the joint angular velocity critical value, a linear amplification factor is used, and the amplification curve is dynamically calibrated through the joint range of motion - virtual model proportional coefficient; a velocity - density correlation function is constructed for the motion speed parameter, a mathematical relationship between the velocity parameter and the virtual obstacle spatial density is established based on the cubic power function, the normalized velocity parameter is input into the Poisson distribution model to generate the obstacle probability distribution field, and the virtual obstacle generation frequency is calculated in real - time according to the density gradient field; an intensity classification logic is set according to the training intensity parameter, and the intensity parameter is divided into three levels: low, medium, and high through the double - threshold segmentation method. The hybrid weights of the multi - modal prompting strategy are determined by combining the fuzzy membership function. The low - intensity level activates visual prompts (color gradient and icon flashing), the medium - intensity level superimposes auditory prompts (frequency - adjustable beeping), and the high - intensity level introduces tactile prompts (vibration intensity and rhythm coding). The collaborative activation coefficient of the multi - modal prompts is dynamically adjusted based on the real - time intensity data flow.

[0143] The exoskeleton is driven and the virtual scene is rendered according to the exoskeleton control instructions and virtual reality scene parameters, and the actual joint torque, trunk attitude angle and fixation point coordinate data are synchronously collected based on the six-axis force sensor array, inertial measurement unit and eye movement tracking module.

[0144] It should be noted that, based on the strain gauge measurement unit of the six-axis force sensor array of the exoskeleton joint, the force components of the X, Y, and Z axes and the torque components around the axes in each joint coordinate system are collected in real time. The original signal is quantized by a 24-bit analog-to-digital converter at a sampling rate of 1000 Hz. After eliminating high-frequency noise by using a second-order Butterworth low-pass filter, the voltage signal is converted into an actual torque vector in Newton-meter units according to the pre-calibrated sensor sensitivity matrix; an inertial measurement unit integrating a three-axis gyroscope, an accelerometer and a magnetometer is used to fuse the angular velocity, linear acceleration and geomagnetic data through a quaternion extended Kalman filter algorithm, and the pitch angle, roll angle and yaw angle of the patient's trunk attitude angle are solved, and the attitude angle Euler angle data is output at a sampling rate of 200 Hz, and the sensor drift error is eliminated through a temperature compensation algorithm; the eye movement tracking module based on the infrared corneal reflection principle uses a 940 nm non-invasive light source and a high-speed CMOS image sensor to capture the eye movement characteristics, calculates the three-dimensional space coordinates of the fixation point by using the pupil-corneal reflection vector method, maps the original eye movement data to the normalized screen coordinate system of the virtual reality scene through a nine-point calibration algorithm, and uses an adaptive threshold filter to eliminate the interference of blink artifacts, and outputs the X-Y two-dimensional time series of the fixation point coordinates.

[0145] S6: Dynamically update the parameters of signal denoising, feature fusion and strategy optimization based on the execution feedback data;

[0146] Furthermore, a torque error scalar is generated through the root mean square error between the actual torque and the expected torque of the exoskeleton joint, an attitude coordination error is calculated based on the dynamic time warping distance of the trunk attitude angle, and a visual attention error is generated by combining the fixation point overlap rate of the virtual task;

[0147] It should be noted that within a 500 - ms time window, 1000 sampling points of the six - dimensional torque of each joint are collected. The mean square of the Euclidean distance between the actual torque vector and the target torque vector is calculated, and after taking the square root, it is normalized to the interval [0, 1]. When the scalar value exceeds 0.5, the parameter update of the noise reduction module is triggered; for the trunk attitude angle data, the dynamic time warping algorithm is used to calculate the attitude coordination error. The time - series sequences of the pitch angle and roll angle collected by the inertial measurement unit are non - rigidly aligned with the ideal rehabilitation trajectory. By restricting the bending window to 20% of the trajectory length, the optimal warping path is constructed, and the cumulative distance value is calculated and divided by the length of the warping path to generate a normalized error index. When the error exceeds 10 degrees, it is determined that the acceleration signal noise reduction fails; based on the virtual task fixation point coordinate data, the visual attention error is generated by calculating the Jaccard overlap rate between the fixation point heat map and the target area. The error coefficient is defined by 1 minus the ratio of the overlapping area to the target area. After smoothing the original eye movement data by combining with the Kalman filter, when the error coefficient exceeds 0.3 for 3 consecutive seconds, the parameter correction of the mental entropy estimation module is triggered.

[0148] Dynamically adjust the wavelet packet sub - band noise threshold of the surface electromyogram signal according to the torque error scalar, correct the R - wave detection noise threshold of the electrocardiogram signal by median filtering based on the attitude coordination error, and update the parameters of signal noise reduction, feature fusion, and strategy optimization by combining the visual attention error;

[0149] It should be noted that when the torque error exceeds the torque error threshold, the response intensity of the hyperbolic tangent function is enhanced by the gain factor, so that the noise threshold increases step - by - step with the increase of the error, suppressing the interference of high - frequency motion artifacts; for the error signal with an amplitude lower than the torque error threshold, the exponential decay mode is used to gradually restore the basic noise suppression level. For the correction of the R - wave detection noise threshold of the electrocardiogram signal, dynamic weighted fusion is carried out by combining the attitude coordination error and the median filtering result: the statistical characteristics of the electrocardiogram signal baseline amplitude are extracted by sliding window median filtering, and the linear compensation term of the attitude error is superimposed. When the attitude instability causes a sharp increase in the error, the threshold emergency floating mechanism is activated to suppress the misdetection of abnormal R - waves. In the process of optimizing the gravity compensation of the acceleration signal, a learning rate adaptive adjustment strategy is constructed based on the visual attention error: the exponential decay function is used to map the visual attention error to the learning rate correction coefficient of the gradient descent algorithm, accelerating the iterative convergence speed of the gravity component in the high - error state, and maintaining the steady - state optimization to prevent over - fitting in the low - error state. The rotation matrix parameters and the gravity vector estimation residuals are associated through chain - rule derivation to improve the dynamic accuracy of the spatial attitude solution.

[0150] The root mean square error between the actual torque and the target torque of the exoskeleton joint is used as the loss function. The partial derivative of the error with respect to the input gate weight matrix is calculated through chain differentiation. The momentum optimizer is used to fuse the historical gradient direction and the current gradient value to generate the weight update amount. When the torque error exceeds the safety threshold for three consecutive frames, the gradient clipping mechanism is activated to limit the update amplitude. The dynamic pruning of the hypergraph adjacency matrix and the decay of hyperedge weights are guided by the visual attention error. The exponential moving average of the attention error is set as the pruning trigger condition. When the error mean exceeds 0.25, the hyperedge set is traversed, and the hyperedges with the phase synchronization value of the associated nodes lower than 0.4 are subjected to the weight coefficient decay operation. The decay factor decreases linearly in segments with the error intensity. The hyperedge connections with the weight continuously lower than 0.1 are synchronously removed. During the sparsification process of the multi-head attention mask, L1 regularization constraints are introduced, an absolute value penalty term is imposed on the dot product result of the query vector and the key vector, and the proximal gradient descent algorithm is used to separate the significant components and noise components in the attention weights, retain the high-weight connections of the nodes associated with the head fixation area, and force the attention weights of non-related node pairs to zero, finally forming a sparsified attention mask pattern that only retains the top 30% significant connections.

[0151] Taking the number of safety events per hour as the input variable, the piecewise linear interpolation method is used to generate the threshold adjustment coefficient. When the event frequency exceeds 3 times per hour, the threshold increasing mode is triggered, and the upper limit of the safety threshold is gradually increased in steps of 0.05. When the event frequency is lower than 1 time per hour, the threshold recovery mechanism is started, and the baseline threshold is gradually regressed through the exponential decay function. During the process of optimizing the exploration decay factor through the policy gradient algorithm, a two-dimensional reward function including the proportion of the new action mode and the safety index is constructed. The Monte Carlo sampling is used to evaluate the policy improvement direction, and the Adam optimizer is combined to calculate the gradient update amount of the exploration decay factor. The learning rate is set to 0.01. When the reward growth rate is less than 2% within 5 consecutive training cycles, the learning rate cosine annealing mechanism is activated to prevent local optima. When adjusting the superposition ratio of the inertia factor in stages, it is divided into three stages: the initial stage (0 - 14 days), the middle stage (15 - 42 days), and the late stage (more than 43 days) according to the rehabilitation training cycle. In the initial stage (0 - 14 days), a fixed ratio of 0.3 is maintained. In the middle stage (15 - 42 days), the exponential decay function is used to dynamically adjust the superposition weight, and the decay time constant is set to 30 days. In the late stage (more than 43 days), the inertia factor ratio is locked at 0.05. Each time the stage is switched, the sensitivity of the ratio parameter to the torque error is verified through the gradient descent algorithm. When the error change rate exceeds 10%, the ratio parameter freeze protection mechanism is triggered.

[0152] This embodiment also provides an accelerated rehabilitation data processing device based on electrophysiological data, including an acquisition module for synchronously acquiring surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals, and generating a time-aligned noise reduction matrix through an adaptive noise reduction algorithm; a fusion module for calculating the full-modal phase-locking value based on the noise reduction matrix and generating a fusion feature vector through a gated recurrent network; a multi-scale entropy state analysis module for extracting micro-entropy, meso-entropy, macro-entropy, and psychological entropy from the fusion feature vector and constructing an entropy state vector; a personalized rehabilitation strategy generation module for generating personalized rehabilitation parameters under safety constraints and exploration constraints according to the entropy state vector; an exoskeleton-VR execution control module for converting the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters and collecting execution feedback data; a dynamic parameter optimization module for dynamically updating the parameters of the signal noise reduction, feature fusion, and strategy optimization modules based on the execution feedback data.

[0153] This embodiment also provides a computer device applicable to the case of an accelerated rehabilitation data processing method based on electrophysiological data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the accelerated rehabilitation data processing method based on electrophysiological data proposed in the above embodiment.

[0154] This computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a carrier network, NFC (near field communication), or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the computer device housing, or an external keyboard, a touchpad, or a mouse, etc.

[0155] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for processing accelerated rehabilitation data based on electrophysiological data proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0156] In summary, the present invention significantly improves the quality and accuracy of multi-modal electrophysiological signal processing through synchronous acquisition and adaptive noise reduction, as well as calculating the full-modal phase-locking value based on the noise reduction matrix and generating a fusion feature vector. The former ensures the consistency and synchronization of signals through precise time alignment and high-quality noise reduction; the latter realizes the optimization of personalized rehabilitation parameters through deep feature extraction and dynamic adjustment mechanisms. These innovations not only solve the key problems in the prior art, but also show broad application prospects in the cross-field of biomedical engineering and artificial intelligence.

[0157] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An accelerated rehabilitation data processing method based on electrophysiological data, characterized in that: including, simultaneously collecting surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals, and generating a time-aligned noise reduction matrix through an adaptive noise reduction algorithm; calculating the full-modal phase-locking value based on the noise reduction matrix, and generating a fused feature vector through a gated recurrent network; extracting micro-entropy, meso-entropy, macro-entropy, and psychological entropy from the fused feature vector, and constructing an entropy state vector; generating personalized rehabilitation parameters under safety constraints and exploration constraints according to the entropy state vector; converting the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters, and collecting execution feedback data; dynamically updating the parameters of signal noise reduction, feature fusion, and strategy optimization based on the execution feedback data.

2. The accelerated rehabilitation data processing method based on electrophysiological data according to claim 1, wherein: The specific steps for generating the time-aligned noise reduction matrix through the adaptive noise reduction algorithm are as follows. Performing noise reduction processing on the collected surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals; Aligning the time through hardware-level synchronization, resampling alignment, matrix construction, and data verification, and generating a noise reduction matrix.

3. The accelerated rehabilitation data processing method based on electrophysiological data according to claim 2, characterized in that: The specific steps for calculating the full-modal phase-locking value based on the noise reduction matrix and generating a fused feature vector through a gated recurrent network are as follows. Extracting the time-frequency phase of the surface electromyogram signal, the piecewise linear phase of the electrocardiogram signal, the exponentially decaying phase of the skin electrical activity signal, and the phase-locked loop tracking phase of the acceleration signal from the noise reduction matrix, and generating a phase matrix by aligning them along the time axis; Calculating the full-modal phase-locking value based on the phase matrix through a multi-modal PLV formula; Defining a node set and a hyperedge set based on the full-modal phase-locking value, and constructing a dynamic hypergraph through a hypergraph adjacency matrix; Allocating the input weights of the signal channels based on the full-modal phase-locking value through the input gate and the forget gate, and adjusting the forgetting ratio of the historical hidden state according to the acceleration phase difference; Combining the input gate weights and the historical hidden state adjusted by the forget gate to generate a candidate hidden state and update the hidden state; Aggregating the hidden state through multi-head attention with hypergraph constraints to generate a fused feature vector.

4. The accelerated rehabilitation data processing method based on electrophysiological data according to claim 3, wherein: The specific steps for extracting micro-entropy, meso-entropy, macro-entropy, and psychological entropy from the fused feature vector and constructing an entropy state vector are as follows. Extracting entropy features from the fused feature vector at the micro, meso, macro, and psychological levels to generate micro-entropy, meso-entropy, macro-entropy, and psychological entropy; Constructing an entropy state vector through the extracted micro-entropy, meso-entropy, macro-entropy, and psychological entropy.

5. The accelerated rehabilitation data processing method based on electrophysiological data according to claim 4, wherein: The specific steps for generating personalized rehabilitation parameters under safety constraints and exploration constraints according to the entropy state vector are as follows. Setting safety thresholds C1, C3, and C4 and a time decay factor based on the historical rehabilitation dataset; Dynamically restricting high-risk actions through safety constraint modeling, including When the micro-entropy exceeds the safety threshold C1, reducing the training intensity parameter through a Sigmoid decay function; When the macro-entropy exceeds the safety threshold C3, restricting the joint activity range parameter through an exponential truncation function; When the psychological entropy exceeds the safety threshold C4, reducing the movement speed parameter through a piecewise linear zeroing function; Encouraging diverse action patterns through exploration constraint modeling, including Generate the exploration weights for training intensity through the hyperbolic tangent function, generate the exploration weights for the range of joint motion through the inverse exponential function, generate the exploration weights for movement speed through the power function, and superimpose the time decay factor; Multiply the training intensity parameters element-wise with the exploration weights for training intensity to generate the preliminary training intensity parameters; Multiply the safety parameters for the range of joint motion element-wise with the exploration weights for the range of joint motion to generate the preliminary parameters for the range of joint motion; Multiply the safety parameters for movement speed element-wise with the exploration weights for movement speed to generate the preliminary movement speed parameters; Superimpose the exploration weights for training intensity, the range of joint motion, and movement speed with the corresponding parameters respectively to form enhanced, adaptable, and driving intermediate parameters, and then perform a secondary superposition to generate the final personalized rehabilitation parameters.

6. The accelerated rehabilitation data processing method based on electrophysiological data according to claim 5, characterized in that: The conversion of the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters, and the acquisition of execution feedback data are specifically as follows: Convert the parameters for the range of joint motion into a spatial coordinate sequence through the cubic spline interpolation algorithm; Convert the training intensity parameters into the motor torque output threshold through the fuzzy PID controller; Convert the movement speed parameters into the joint angular velocity constraint values through the inverse solution algorithm of the Jacobian matrix; Use the spatial coordinate sequence, the motor torque output threshold, and the joint angular velocity constraint values as the exoskeleton control instructions; Convert the parameters for the range of joint motion into the motion amplitude of the virtual limb model through the non-linear amplification mapping rule; Convert the movement speed parameters into the virtual obstacle generation frequency through the velocity-density correlation function; Convert the training intensity parameters into a multi-modal prompting strategy through the intensity classification logic, and use the motion amplitude of the virtual limb model, the virtual obstacle generation frequency, and the multi-modal prompting strategy as the virtual reality scene parameters; Perform exoskeleton drive execution and virtual scene rendering through the exoskeleton control instructions and the virtual reality scene parameters; Synchronously collect the actual joint torque, trunk attitude angle, and fixation point coordinate data based on the six-dimensional force sensor array, inertial measurement unit, and eye movement tracking module.

7. The accelerated rehabilitation data processing method based on electrophysiological data according to claim 6, wherein: The update of the parameters for signal denoising, feature fusion, and strategy optimization based on the execution feedback data is specifically as follows: Generate the torque error scalar through the root mean square error between the actual joint torque of the exoskeleton and the expected torque, calculate the attitude coordination error based on the dynamic time warping distance of the trunk attitude angle, and generate the visual attention error in combination with the fixation point overlap rate of the virtual task; Dynamically adjust the wavelet packet sub-band noise threshold of the surface electromyogram signal according to the torque error scalar, perform median filtering correction on the R-wave detection noise threshold of the electrocardiogram signal based on the attitude coordination error, and update the parameters for signal denoising, feature fusion, and strategy optimization in combination with the visual attention error.

8. An accelerated rehabilitation data processing device based on electrophysiological data, based on the accelerated rehabilitation data processing method based on electrophysiological data according to any one of claims 1 to 7, characterized in that: It includes an acquisition module, a fusion module, a multi-scale entropy state analysis module, a personalized rehabilitation strategy generation module, an exoskeleton-VR execution control module, and a dynamic parameter optimization module. The acquisition module is used to synchronously collect surface electromyogram signals, electrocardiogram signals, skin electrical activity signals, and acceleration signals, and generate a time-aligned denoising matrix through the adaptive denoising algorithm; The fusion module is used to calculate the full-modal phase locking value based on the denoising matrix and generate a fusion feature vector through the gated recurrent network; The multi-scale entropy state analysis module is used to extract micro entropy, mesoscopic entropy, macro entropy and psychological entropy from the fusion feature vector, and construct an entropy state vector; The personalized rehabilitation strategy generation module is used to generate personalized rehabilitation parameters according to the entropy state vector under safety constraints and exploration constraints; The exoskeleton-VR execution control module is used to convert the personalized rehabilitation parameters into exoskeleton control instructions and virtual reality scene parameters, and collect execution feedback data; The dynamic parameter optimization module is used to dynamically update the parameters of the signal denoising, feature fusion and strategy optimization modules based on the execution feedback data.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the accelerated rehabilitation data processing method based on electrophysiological data according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the accelerated rehabilitation data processing method based on electrophysiological data according to any one of claims 1 to 7.

Images