Muscle stimulation adjusting method and system based on gait detection

Abstract

The invention provides a muscle stimulation adjusting method and system based on gait detection. The method comprises the following steps: acquiring lower limb biomechanical data; performing phase recognition according to a dynamic programming algorithm, and constructing a pressure center track to perform feature extraction to obtain gait features; carrying out feature analysis on the gait features through a reverse dynamics model and a machine learning model, and carrying out output judgment according to a double-model collaborative decision-making mechanism to obtain a stimulation parameter decision-making result; generating an electrical stimulation waveform according to the decision result, and applying differential electrical stimulation to the target muscle in the gait cycle to obtain a stimulation scheme; and monitoring stimulated myoelectricity feedback signals and gait changes, and carrying out parameter updating by calculating a signal change rate and analyzing a motion track error to obtain an updating scheme. According to the method, through a double decision-making mechanism of the reverse dynamics model and the machine learning model and a real-time feedback regulation mechanism, the gait phase recognition precision is improved, and precise regulation and control of stimulation parameters are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical stimulation, and particularly to a muscle stimulation regulation method and system based on gait detection. Background Art

[0002] Functional electrical stimulation (FES) technology is now widely used to improve the gait performance of patients with movement disorders. Traditional FES systems usually adopt a pre-programmed fixed stimulation pattern, and apply electrical pulses to target muscles through surface electrodes to induce muscle contraction and limb movement. In recent years, with the development of wearable sensor technology, adaptive FES systems based on gait detection have gradually become a research hotspot.

[0003] However, the traditional fixed stimulation pattern is difficult to adapt to individual differences and dynamically changing gait characteristics, which may lead to problems such as muscle fatigue and gait abnormalities; and the decision-making of stimulation parameters mostly relies on a single model, resulting in insufficient matching degree between the stimulation scheme and the actual needs. Therefore, it is very necessary to design a muscle stimulation regulation method and system based on gait detection. Summary of the Invention

[0004] The purpose of the present invention is to provide a muscle stimulation regulation method and system based on gait detection, which can improve the accuracy of gait phase recognition and achieve precise regulation of stimulation parameters through a dual decision-making mechanism and a real-time feedback regulation mechanism of an inverse dynamics model and a machine learning model.

[0005] To achieve the above purpose, the present invention provides the following solutions:

[0006] A muscle stimulation regulation method based on gait detection, comprising the following steps:

[0007] Real-time collect the lower limb biomechanical data of the subject through a multi-source sensor group of a wearable device;

[0008] Perform real-time phase recognition on the lower limb biomechanical data according to the dynamic programming algorithm, and construct a center of pressure trajectory to extract features from the recognition results to obtain gait features;

[0009] Respectively perform feature analysis on the gait features through the constructed inverse dynamics model and machine learning model, and perform output determination on the analysis results according to the dual-model collaborative decision-making mechanism to obtain a stimulation parameter decision result;

[0010] Generate an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and apply differential electrical stimulation to the target muscle within the gait cycle through the electrical stimulation waveform to obtain a stimulation scheme;

[0011] Monitor the electromyographic feedback signal and gait changes after applying the stimulation scheme, update the parameters of the stimulation scheme by calculating the signal change rate and analyzing the motion trajectory error, and obtain the updated scheme.

[0012] Optionally, the multi-source sensor group includes: a plantar pressure sensor array, an inertial measurement unit, and a surface electromyographic sensor; the lower limb biomechanical data includes: plantar pressure distribution time series data, three-dimensional lower limb kinematic data, and electromyographic signals of the target muscle group.

[0013] Optionally, perform real-time phase recognition on the lower limb biomechanical data according to the dynamic programming algorithm, and construct the center of pressure trajectory to extract features from the recognition results to obtain gait features, including:

[0014] Eliminate the high-frequency noise in the lower limb biomechanical data through a Butterworth low-pass filter to obtain denoised data;

[0015] Take the first derivative of the angular velocity signal in the denoised data as the observation sequence, and divide the gait state according to the observation sequence;

[0016] Based on the observation sequence, obtain the phase transition time points of the denoised data through the Viterbi algorithm;

[0017] Calculate the pressure-time integral of the gait state, and calculate the phase features at the phase transition time points in combination with the center of pressure trajectory;

[0018] Decompose the phase features through wavelet transform to obtain gait features.

[0019] Optionally, perform feature analysis on the gait features through the constructed inverse dynamics model and machine learning model respectively, and make an output determination on the analysis results according to the dual-model collaborative decision-making mechanism to obtain the stimulation parameter decision result, including:

[0020] Calculate the ankle joint moment of the gait features based on the Newton-Euler equation of the inverse dynamics model, and combine the muscle moment arm parameters to obtain the activation threshold of the target muscle group;

[0021] Calculate the correlation coefficient between the lower limb biomechanical data and the activation threshold of the target muscle group, and adjust the parameters of the inverse dynamics model through the correlation coefficient to obtain the adjusted inverse dynamics model;

[0022] Establish a mapping relationship of the gait features through a long short-term memory network model;

[0023] Use the historical gait data and ideal stimulation parameters as the training set to train the long short-term memory network model to obtain a prediction model;

[0024] The decision result of the stimulation parameters is obtained according to the comparison result between the output difference of the adjusted inverse dynamics model and the predicted model and the preset difference threshold.

[0025] Optionally, calculate the correlation coefficient between the lower limb biomechanical data and the activation threshold of the target muscle group, and adjust the parameters of the inverse dynamics model through the correlation coefficient to obtain the adjusted inverse dynamics model, including:

[0026] Calculate the correlation coefficient according to the time domain synchronization coefficient, frequency domain coupling degree and phase matching degree of the lower limb biomechanical data and the activation threshold of the target muscle group;

[0027] When the correlation coefficient is lower than the preset correlation threshold, iterate and update the inverse dynamics model through a hierarchical progressive mechanism;

[0028] Verify the conservation of momentum in the iterative update process, and perform iterative backtracking when the energy error is greater than the preset loss threshold until the physiological rationality boundary condition is satisfied and the iteration stops, obtaining the adjusted inverse dynamics model.

[0029] Optionally, the calculation formula of the correlation coefficient is: where α, β and γ are the time domain weight factor, frequency domain weight factor and phase compensation coefficient respectively, E t is the EMG RMS value at time t, μ E is the mean value of the EMG signal, τ t is the theoretical value of the ankle joint moment at time t, μ τ is the mean value of the theoretical moment value, T is the time window length, W E (k) is the energy of the kth frequency band of the EMG signal, W τ (k) is the energy of the kth frequency band of the moment signal, k is the frequency band index, K is the total number of frequency bands, θ E (t n ) is the instantaneous phase angle of the EMG signal at time t, θ τ (t n ) is the instantaneous phase angle of the moment signal at time t, and N is the number of phase sampling points.

[0030] Optionally, the hierarchical progressive mechanism includes:

[0031] Convert the lower limb biomechanical data into a standard vector;

[0032] Construct a dynamic error matrix according to the standard vector;

[0033] Based on the dynamic error matrix, combine the coupled gradient descent method to iteratively update the inverse dynamics model.

[0034] Optionally, an electrical stimulation waveform with phase synchronization characteristics is generated according to the stimulation parameter decision result, and differential electrical stimulation is applied to the target muscle within the gait cycle through the electrical stimulation waveform to obtain a stimulation plan, including:

[0035] Select a biphasic asymmetric square wave from a preset parameter library as the basic waveform according to the stimulation parameter decision result;

[0036] Activate the stimulation of the gastrocnemius muscle after a delay time at the phase transition time point within the gait cycle to obtain the gastrocnemius muscle stimulation intensity;

[0037] Activate the stimulation of the tibialis anterior muscle when the anterior tibial angle within the gait cycle reaches the mid-swing threshold to obtain the tibialis anterior muscle stimulation intensity;

[0038] Calculate the compensation coefficients of the gastrocnemius muscle stimulation intensity and the tibialis anterior muscle stimulation intensity according to the electrode-skin impedance, and adaptively compensate the gastrocnemius muscle stimulation intensity and the tibialis anterior muscle stimulation intensity through the compensation coefficients to obtain the gastrocnemius muscle compensation intensity and the tibialis anterior muscle compensation intensity;

[0039] Calculate the muscle stimulation interval through the walking speed, and perform interleaved stimulation on the gastrocnemius muscle and the tibialis anterior muscle according to the muscle stimulation interval to obtain a stimulation plan.

[0040] Optionally, monitor the electromyographic feedback signal and gait changes after applying the stimulation plan, and update the parameters of the stimulation plan through calculating the signal change rate and analyzing the motion trajectory error to obtain an updated plan, including:

[0041] Calculate the root mean square and median frequency of the electromyographic feedback signal, and judge the muscle state according to the root mean square and median frequency;

[0042] Obtain the ground contact impact rate according to the gait changes, and calculate the buffering efficiency index according to the ground contact impact rate;

[0043] Construct the lower limb motion trajectory according to the gait changes, and calculate the mean square error between the joint angle and the standard gait according to the lower limb motion trajectory;

[0044] Adjust the parameters in the stimulation plan according to the muscle state, buffering efficiency index and mean square error to obtain an updated plan.

[0045] A muscle stimulation regulation system based on gait detection, including:

[0046] A gait detection module, configured to collect the lower limb biomechanical data of the subject in real time through a multi-source sensor group of a wearable device;

[0047] A gait recognition module, configured to perform real-time phase recognition on the lower limb biomechanical data according to a dynamic programming algorithm, and construct a center of pressure trajectory to extract features of the recognition result to obtain gait features;

[0048] The gait analysis module is used to analyze the gait features through the constructed inverse dynamics model and machine learning model respectively, and output and determine the analysis results according to the dual-model collaborative decision-making mechanism to obtain the stimulation parameter decision result;

[0049] The stimulation scheme generation module is used to generate an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and apply differential electrical stimulation to the target muscle within the gait cycle through the electrical stimulation waveform to obtain the stimulation scheme;

[0050] The feedback regulation module is used to monitor the electromyogram feedback signal and gait changes after applying the stimulation scheme, and update the parameters of the stimulation scheme through calculating the signal change rate and analyzing the movement trajectory error to obtain the updated scheme.

[0051] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: The muscle stimulation regulation method and system based on gait detection provided by the present invention, the method includes: real-time collecting the lower limb biomechanical data of the subject through the multi-source sensor group of the wearable device; performing real-time phase recognition on the lower limb biomechanical data according to the dynamic programming algorithm, and constructing the center of pressure trajectory to extract features of the recognition result to obtain gait features; analyzing the gait features through the constructed inverse dynamics model and machine learning model respectively, and outputting and determining the analysis results according to the dual-model collaborative decision-making mechanism to obtain the stimulation parameter decision result; generating an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and applying differential electrical stimulation to the target muscle within the gait cycle through the electrical stimulation waveform to obtain the stimulation scheme; monitoring the electromyogram feedback signal and gait changes after applying the stimulation scheme, and updating the parameters of the stimulation scheme through calculating the signal change rate and analyzing the movement trajectory error to obtain the updated scheme. This method improves the gait phase recognition accuracy through the dual decision-making mechanism of the inverse dynamics model and the machine learning model and the real-time feedback regulation mechanism, and realizes the precise control of the stimulation parameters. Description of the Drawings

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the 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 drawings can also be obtained based on these drawings.

[0053] Figure 1 It is the flowchart of the muscle stimulation regulation method of the present invention;

[0054] Figure 2 It is the flowchart of the gait feature extraction of the present invention;

[0055] Figure 3 It is the flowchart for judging the decision result of the stimulation parameters of the present invention;

[0056] Figure 4 It is the flowchart for generating the stimulation scheme of the present invention;

[0057] Figure 5 It is the flowchart for updating the stimulation scheme of the present invention. Specific embodiments

[0058] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0059] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] As Figure 1 shown, the present invention provides a muscle stimulation regulation method based on gait detection, including the following steps:

[0061] Step 100: Real-time collect the lower limb biomechanical data of the subject through the multi-source sensor group of the wearable device;

[0062] Specifically, the multi-source sensor group includes: a plantar pressure sensor array, an inertial measurement unit, and a surface electromyography sensor. The plantar pressure sensor array is distributed in a 5×8 matrix form, with a sampling frequency ≥200Hz, and is used to detect the pressure distribution and change time series of 40 regions of the foot in real time. The inertial measurement unit (IMU) includes a three-axis accelerometer and a three-axis gyroscope. The IMU is fixed to the arch of the foot and collects three-dimensional acceleration and angular velocity data at a frequency of 100Hz. The surface electromyography sensor adopts a differential electrode configuration and is attached to the tibialis anterior muscle and the lateral head of the gastrocnemius muscle. Its bandwidth is 10 - 500Hz, and the sampling rate is 2000Hz. The lower limb biomechanical data includes: plantar pressure distribution time series data, three-dimensional lower limb kinematic data, and electromyography signals of the target muscle group.

[0063] Step 200: Perform real-time phase recognition on the lower limb biomechanical data according to the dynamic programming algorithm, and construct a center of pressure trajectory to extract features from the recognition result to obtain gait features; The specific steps are as Figure 2 shown, including:

[0064] Step 201: Eliminate the high-frequency noise of the lower limb biomechanical data through a Butterworth low-pass filter to obtain denoised data;

[0065] Specifically, a fourth-order Butterworth low-pass filter with a cut-off frequency of 15 Hz is used to filter out high-frequency interference caused by sensor electronic noise, muscle tremors, etc., while retaining the key biomechanical features related to gait. The zero-phase bidirectional filtering technique is adopted in the filtering process to avoid phase distortion caused by signal time shift. The plantar pressure sensor array adopts a parallel processing architecture, and each channel of the array independently performs filtering operations, which not only ensures the real-time nature of the data but also improves the signal-to-noise ratio.

[0066] Step 202: Use the first derivative of the angular velocity signal in the denoised data as the observation sequence, and divide the gait state according to the observation sequence;

[0067] Specifically, after the sagittal plane angular velocity signal obtained by the IMU is filtered, the central difference method is used to calculate its first derivative and used as the observation sequence. The gait state is divided by the zero-crossing points, extreme points, and amplitude changes of the observation sequence. The implementation process in this embodiment includes: when the first derivative exceeds the positive threshold of +500 ° / s 2 the current gait state is marked as a heel strike event; when the first derivative is lower than the negative threshold of -300 ° / s 2 the current gait state is marked as a toe-off event.

[0068] Step 203: Based on the observation sequence, obtain the phase transition time points of the denoised data through the Viterbi algorithm;

[0069] Specifically, the observation sequence is input into the hidden Markov model, and the optimal state sequence is solved through the built-in Viterbi algorithm of the model, so as to obtain the phase transition time points. The phase transition time points are the transition points between the four gait states of heel strike (HS), flat foot support (FF), toe off (TO), and swing phase (SW). The formula for the solution process is:

[0070] δ t (j) = max[δ t-1 (i)a ij b j (o t ) ;

[0071] where a ij is the state transition probability, b j (o t ) is the observation probability of state j, δ t (j) is the optimal path probability of state j at time t, δ t-1 (i) is the optimal path probability of state i at time t-1, and o t is the observation value at time t.

[0072] It should be noted that through the global optimal path search of the Viterbi algorithm, the misjudgment problem caused by instantaneous noise interference is overcome, and the detection accuracy is improved.

[0073] Step 204: Calculate the pressure-time integral (PTI) of the gait state, and calculate the phase characteristics at the phase transition time point in combination with the center of pressure (COP) trajectory;

[0074] Specifically, PTI realizes the detection of abnormal force patterns such as arch collapse through the quantification of the load distribution characteristics of each region of the foot. And the dynamic characteristics of the COP trajectory have a strong correlation with ankle joint stability, thus realizing the visualization and quantitative evaluation of gait mechanical characteristics.

[0075] Step 205: Decompose the phase characteristics through wavelet transform to obtain gait characteristics.

[0076] Specifically, use the 4-wavelet basis to perform 5-layer wavelet decomposition on the COP trajectory, extract the energy characteristics in the frequency bands of 0.5 - 2 Hz and 2 - 4 Hz respectively, calculate the proportion of the energy of each frequency band of the energy characteristics, and at the same time analyze the instantaneous frequency of the COP trajectory through the wavelet ridge line, calculate its deviation from the theoretical step frequency, and finally obtain a 128-dimensional time-frequency feature vector, which is regarded as gait characteristics.

[0077] Step 300: Perform feature analysis on the gait characteristics through the constructed inverse dynamics model and machine learning model respectively, and make an output decision on the analysis results according to the dual-model collaborative decision-making mechanism to obtain the decision result of the stimulation parameters; The specific steps are as Figure 3 shown, including:

[0078] Step 301: Calculate the ankle joint moment of the gait characteristics based on the Newton-Euler equation of the inverse dynamics model, and combine the muscle moment arm parameters to obtain the activation threshold of the target muscle group;

[0079] Specifically, simplify the foot, calf, and thigh into a rigid body link system, and then calculate the ankle joint moment through the formula L = I·a + I·ω + ∑(W i ·F i ), where L is the ankle joint moment, I is the moment of inertia matrix, a is the angular acceleration, ω is the angular velocity, W i is the position vector of the i-th acting point of the plantar pressure, and F i is the three-dimensional force measured by the i-th pressure sensor. Then obtain the theoretical activation threshold through the ratio of the muscle moment arm parameter and the angle between the muscle pulling line and the joint rotation axis. Normalize the theoretical activation threshold to a percentage of the maximum voluntary contraction to obtain the activation threshold of the target muscle group.

[0080] Step 302: Calculate the correlation coefficient between the lower limb biomechanical data and the activation threshold of the target muscle group, and adjust the parameters of the inverse dynamics model through the correlation coefficient to obtain the adjusted inverse dynamics model; specifically including:

[0081] Calculate the correlation coefficient based on the time-domain synchronization coefficient, frequency-domain coupling degree, and phase matching degree between the lower limb biomechanical data and the activation threshold of the target muscle group; the calculation formula for the correlation coefficient is:

[0082]

[0083] where α, β, and γ are the time-domain weight factor, frequency-domain weight factor, and phase compensation coefficient respectively, E t is the EMG RMS value at time t, μ E is the mean value of the EMG signal, τ t is the theoretical value of the ankle joint moment at time t, μ τ is the mean value of the theoretical moment value, T is the time window length, W E (k) is the energy of the k-th frequency band of the EMG signal, W τ (k) is the energy of the k-th frequency band of the moment signal, k is the frequency band index, K is the total number of frequency bands, θ E (t n ) is the instantaneous phase angle of the EMG signal at time t, θ τ (t n ) is the instantaneous phase angle of the moment signal at time t, N is the number of phase sampling points. The time-domain weight factor is calculated through the step frequency, and the calculation formula is: The frequency-domain weight factor is obtained by comparing the peak foot pressure and the pressure rise slope, and the calculation formula is: The phase compensation coefficient is 0.4 in this embodiment.

[0084] When the correlation coefficient is lower than the preset correlation threshold, the inverse dynamics model is iteratively updated through a hierarchical progressive mechanism;

[0085] Verify the conservation of momentum during the iterative update process, and perform iterative backtracking when the energy error is greater than the preset loss threshold until the physiological rationality boundary condition is satisfied and the iteration stops, obtaining the adjusted inverse dynamics model.

[0086] Furthermore, the hierarchical progressive mechanism is: convert the lower limb biomechanical data into a standard vector; construct a dynamic error matrix based on the standard vector; and iteratively update the inverse dynamics model based on the dynamic error matrix in combination with the coupled gradient descent method.

[0087] Step 303: Establish a mapping relationship of gait characteristics through a long short-term memory network model;

[0088] Specifically, the input layer of the long short-term memory network receives a 128-dimensional gait feature vector, the hidden layer has 64 memory units, and the output layer is used to generate 4-dimensional stimulation parameters (intensity, frequency, pulse width, and timing). The network uses the sigmoid activation function as the forgetting gate.

[0089] Step 304: Use the historical gait data and ideal stimulation parameters as a training set to train the long short-term memory network model to obtain a prediction model;

[0090] Step 305: Obtain the stimulation parameter decision result based on the comparison result between the output difference of the adjusted inverse dynamics model and the preset difference threshold.

[0091] Specifically, the relative difference degree between the outputs of the two models is calculated by a formula. If the relative difference degree ≤ 15%, the weighted average of the two outputs is taken as the stimulation parameter decision result; if 15% < relative difference degree ≤ 25%, the output closer to the electromyogram feedback is selected from the expert database constructed based on historical data as the stimulation parameter decision result; if the relative difference degree > 25%, these two outputs are excluded and the historical optimal output is selected as the stimulation parameter decision result.

[0092] Step 400: Generate an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and apply differential electrical stimulation to the target muscle within the gait cycle to obtain a stimulation plan;

[0093] The specific steps are as Figure 4 shown, including:

[0094] Step 401: Select a biphasic asymmetric square wave as the basic waveform from the preset parameter library according to the stimulation parameter decision result;

[0095] Specifically, the anode phase pulse width of the biphasic asymmetric square wave is 1.5 times that of the cathode phase, the basic frequency range is 20 - 100 Hz, and the pulse width range is 100 - 400 μs.

[0096] Step 402: Activate the stimulation of the gastrocnemius muscle after a delay time at the phase transition moment point within the gait cycle to obtain the gastrocnemius muscle stimulation intensity;

[0097] Specifically, the calculation formula for the delay time is: where, T d is the delay time. The dynamic acquisition mechanism of the delay time can adapt to the nerve conduction time difference at different gait frequencies, greatly reducing the error of the gastrocnemius muscle activation moment.

[0098] Step 403: Activate the stimulation of the tibialis anterior muscle when the anterior tibial angle within the gait cycle reaches the mid-swing threshold to obtain the tibialis anterior muscle stimulation intensity;

[0099] Specifically, the mid-swing threshold in this embodiment is 15°, the floating range is ±3°, and the muscle stimulation is performed using a burst mode.

[0100] Step 404: Calculate the compensation coefficients for the gastrocnemius muscle stimulation intensity and the tibialis anterior muscle stimulation intensity according to the electrode-skin impedance, and adaptively compensate the gastrocnemius muscle stimulation intensity and the tibialis anterior muscle stimulation intensity through the compensation coefficients to obtain the compensated gastrocnemius muscle intensity and the compensated tibialis anterior muscle intensity;

[0101] Specifically, the calculation formula for the compensation coefficient is: where C is the compensation coefficient, Z b is the initial impedance value, and Z C is the AC impedance value of the current electrode-skin contact interface. The compensated intensity includes voltage compensated intensity and pulse width compensated intensity, and the calculation formulas are respectively: V a = V s ·C; M a = M s ·(1 + 0.15·ln C); where V a is the voltage compensated intensity, V s is the reference voltage, M a is the pulse width compensated intensity, and M s is the reference pulse width.

[0102] It should be noted that voltage compensation can quickly respond to impedance changes and maintain the target current, while pulse width compensation optimizes the charge transfer efficiency through non-linear adjustment, solving the problem of unstable output caused by skin impedance changes in the electrical stimulation system.

[0103] Step 405: Calculate the muscle stimulation interval through the step frequency, and alternately stimulate the gastrocnemius muscle and the tibialis anterior muscle according to the muscle stimulation interval to obtain a stimulation scheme.

[0104] Specifically, the muscle stimulation interval is 0.4 times the step frequency. After the gastrocnemius muscle stimulation ends, trigger the tibialis anterior muscle stimulation after delaying the muscle stimulation interval, and adopt a three-slot interleaved output strategy: distribute the total stimulation current to three 1-ms-wide time slots for transmission, with a time slot interval of 2 ms, and dynamically adjust the pulse density according to the step frequency. When the step frequency > 1.5 Hz, the number of pulses increases from 3 to 5; when the step frequency < 1.0 Hz, the number of pulses decreases to 2.

[0105] Step 500: Monitor the electromyogram feedback signal and gait changes after applying the stimulation scheme, and update the parameters of the stimulation scheme through calculating the signal change rate and analyzing the motion trajectory error to obtain an updated scheme.

[0106] The specific steps are as Figure 5 shown, including:

[0107] Step 501: Calculate the root mean square (RMS) and median frequency (MF) of the electromyography feedback signal, and judge the muscle state according to the RMS and MF;

[0108] Specifically, when the decrease amplitude of RMS > 15% and the left shift amplitude of MF > 10%, it is determined that the muscle state is muscle fatigue; when the increase amplitude of RMS < 5% and MF has no obvious change, it is determined that the muscle state is insufficient stimulation; when the fluctuation amplitude of RMS < 10% and MF is stable, it is determined that the muscle state is good muscle activation.

[0109] Step 502: Obtain the touchdown impact rate according to the gait change, and calculate the buffering efficiency index according to the touchdown impact rate;

[0110] Specifically, the touchdown impact rate is the ratio of the difference between the peak pressure of the plantar pressure sensor and the baseline pressure before touchdown and the difference between the pressure peak moment and the touchdown moment.

[0111] Step 503: Construct the lower limb movement trajectory according to the gait change, and calculate the mean square error between the joint angle and the standard gait according to the lower limb movement trajectory;

[0112] Specifically, based on the measurement data of the IMU, the lower limb movement trajectory including the real-time angles of the hip, knee and ankle joints is calculated by the quaternion fusion algorithm. A standard gait template including the joint angle curves in the sagittal plane, coronal plane and transverse plane is constructed according to the healthy population database. Calculate the mean square error (MSE) between the real-time angle and the standard gait template. When MSE < 5°, the gait is normal; when 5° ≤ MSE < 10°, the gait is mildly abnormal; when MSE ≥ 10°, the gait is significantly abnormal.

[0113] Step 504: Adjust the parameters in the stimulation scheme according to the muscle state, buffering efficiency index and mean square error to obtain an updated scheme.

[0114] Specifically, in this embodiment, a fuzzy logic controller is used for parameter update.

[0115] The present invention also provides a muscle stimulation regulation system based on gait detection, including:

[0116] A gait detection module for real-time collecting the lower limb biomechanical data of the subject through the multi-source sensor group of the wearable device;

[0117] A gait recognition module for real-time phase recognition of the lower limb biomechanical data according to the dynamic programming algorithm, and constructing a center of pressure trajectory to extract features of the recognition result to obtain gait features;

[0118] The gait analysis module is used to analyze the gait features through the constructed inverse dynamics model and machine learning model respectively, and output and determine the analysis results according to the dual-model collaborative decision-making mechanism to obtain the decision result of the stimulation parameters;

[0119] The stimulation scheme generation module is used to generate an electrical stimulation waveform with phase synchronization characteristics according to the decision result of the stimulation parameters, and apply differential electrical stimulation to the target muscle within the gait cycle through the electrical stimulation waveform to obtain the stimulation scheme;

[0120] The feedback regulation module is used to monitor the electromyographic feedback signal and gait changes after applying the stimulation scheme, and update the parameters of the stimulation scheme by calculating the signal change rate and analyzing the motion trajectory error to obtain the updated scheme.

[0121] The beneficial effects of the present invention are as follows:

[0122] 1) Through the collaborative work of the sensor array, inertial measurement unit (IMU) and surface electromyographic sensor, the full-dimensional acquisition of lower limb biomechanical data is realized;

[0123] 2) Through the dual verification mechanism of the inverse dynamics model and the long short-term memory network model, combined with the correlation coefficient arbitration strategy, the decision error of the stimulation parameters is reduced, and the accuracy of obtaining the stimulation intensity is improved;

[0124] 3) Through the adaptive compensation mechanism of electrode-skin impedance monitoring and the collaborative adjustment strategy of voltage and pulse width, the fluctuation of the stimulation current is greatly reduced, and the stability of the stimulation is ensured;

[0125] 4) By triggering differential stimulation through gait phase recognition, the muscle activation efficiency is greatly improved.

[0126] In this specification, each embodiment is described in a progressive manner. The key points of each embodiment are the differences from other embodiments. The same and similar parts between the embodiments can be referred to each other.

[0127] Specific examples are applied in the present invention to elaborate on the principle and implementation manner of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A muscle stimulation adjustment method based on gait detection, characterized in that: The steps include: The biomechanical data of the subjects' lower limbs are collected in real time through a multi-source sensor group of the wearable device; Performing real-time phase recognition on the lower limb biomechanical data according to a dynamic programming algorithm, and constructing a pressure center trajectory to extract features from the recognition results to obtain gait features; The gait characteristics are analyzed respectively by the constructed inverse dynamics model and the machine learning model, and the analysis results are output and determined according to the dual-model collaborative decision-making mechanism to obtain the stimulation parameter decision result; generating an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and applying differentiated electrical stimulation to the target muscle during the gait cycle through the electrical stimulation waveform to obtain a stimulation scheme; The myoelectric feedback signal and gait changes after the stimulation scheme is applied are monitored, and the parameters of the stimulation scheme are updated by calculating the signal change rate and analyzing the motion trajectory error to obtain an updated scheme.

2. The muscle stimulation adjustment method based on gait detection according to claim 1, characterized in that: The multi-source sensor group includes: a plantar pressure sensor array, an inertial measurement unit and a surface electromyography sensor; the lower limb biomechanics data includes: plantar pressure distribution time series data, lower limb kinematics three-dimensional data and electromyography signals of target muscle groups.

3. The muscle stimulation adjustment method based on gait detection according to claim 1, characterized in that: The lower limb biomechanical data is subjected to real-time phase recognition according to a dynamic programming algorithm, and a pressure center trajectory is constructed to extract features of the recognition results to obtain gait features, including: Eliminating high-frequency noise of the lower limb biomechanical data by using a Butterworth low-pass filter to obtain denoised data; taking the first-order derivative of the angular velocity signal in the denoised data as an observation sequence, and dividing the gait state according to the observation sequence; Based on the observation sequence, obtaining the phase conversion time point of the denoised data by using the Viterbi algorithm; Calculating the pressure-time integral of the gait state, and calculating the phase feature at the phase transition time point in combination with the pressure center trajectory; The phase feature is decomposed by wavelet transform to obtain the gait feature.

4. The muscle stimulation adjustment method based on gait detection according to claim 1, characterized in that: The gait characteristics are analyzed by the constructed inverse dynamics model and the machine learning model respectively, and the analysis results are output and determined according to the dual-model collaborative decision-making mechanism to obtain the stimulation parameter decision results, including: The ankle joint torque of the gait characteristic is calculated based on the Newton-Euler equation of the inverse dynamics model, and the activation threshold of the target muscle group is obtained in combination with the muscle moment arm parameter; Calculating the correlation coefficient between the lower limb biomechanical data and the activation threshold of the target muscle group, and adjusting the parameters of the inverse dynamics model according to the correlation coefficient to obtain an adjusted inverse dynamics model; Establishing a mapping relationship of the gait features through a long short-term memory network model; Using historical gait data and ideal stimulation parameters as training sets to train the long short-term memory network model to obtain a prediction model; The stimulation parameter decision result is obtained according to a comparison result between the output difference of the adjusted inverse dynamics model and the prediction model and a preset difference threshold.

5. The muscle stimulation and regulation method based on gait detection according to claim 4, characterized in that: Calculating the correlation coefficient between the lower limb biomechanical data and the target muscle group activation threshold, and adjusting the parameters of the inverse dynamics model according to the correlation coefficient to obtain an adjusted inverse dynamics model, including: The correlation coefficient is calculated based on the time domain synchronization coefficient, frequency domain coupling degree and phase matching degree of the lower limb biomechanics data and the activation threshold of the target muscle group; When the correlation coefficient is lower than a preset correlation threshold, iteratively updating the inverse dynamics model through a hierarchical progressive mechanism; The momentum conservation is verified during the iterative update process. When the energy error is greater than a preset loss threshold, iterative backtracking is performed until the iteration is stopped when the physiological rationality boundary condition is met, thereby obtaining the adjusted inverse dynamics model.

6. The muscle stimulation and regulation method based on gait detection according to claim 5, characterized in that: The calculation formula of the correlation coefficient is: Among them, α, β and γ are the time domain weight factor, frequency domain weight factor and phase compensation coefficient respectively, E t is the electromyography RMS value at time t, μ E is the mean value of the electromyographic signal, τ t is the theoretical value of ankle joint torque at time t, μ τ is the theoretical mean value of the moment, T is the time window length, W E (k) is the energy of the kth frequency band of the electromyographic signal, W τ (k) is the kth frequency band energy of the torque signal, k is the frequency band index, K is the total number of frequency bands, θ E (t n ) is the instantaneous phase angle of the electromyographic signal at time t, θ τ (t n ) is the instantaneous phase angle of the torque signal at time t, and N is the number of phase sampling points.

7. The muscle stimulation and regulation method based on gait detection according to claim 5, characterized in that: The hierarchical progressive mechanism includes: converting the lower limb biomechanical data into a standard vector; Constructing a dynamic error matrix according to the standard vector; Based on the dynamic error matrix, the inverse dynamics model is iteratively updated in combination with a coupled gradient descent method.

8. The muscle stimulation and regulation method based on gait detection according to claim 3, characterized in that: Generating an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and applying differentiated electrical stimulation to the target muscle during the gait cycle through the electrical stimulation waveform to obtain a stimulation scheme, including: Selecting a biphasic asymmetric square wave as a basic waveform from a preset parameter library according to the stimulation parameter decision result; activating stimulation of the gastrocnemius muscle after a delay time at the phase conversion time point in the gait cycle to obtain a gastrocnemius muscle stimulation intensity; When the tibial anterior tilt angle in the gait cycle reaches a mid-swing threshold, activating stimulation of the tibialis anterior muscle to obtain a tibialis anterior muscle stimulation intensity; Calculating compensation coefficients of the gastrocnemius muscle stimulation intensity and the tibialis anterior muscle stimulation intensity according to the electrode-skin impedance, and adaptively compensating the gastrocnemius muscle stimulation intensity and the tibialis anterior muscle stimulation intensity by using the compensation coefficients to obtain gastrocnemius muscle compensation intensity and tibialis anterior muscle compensation intensity; The muscle stimulation interval is calculated by the step frequency, and the gastrocnemius muscle and the tibialis anterior muscle are alternately stimulated according to the muscle stimulation interval to obtain the stimulation scheme.

9. The muscle stimulation and regulation method based on gait detection according to claim 1, characterized in that: Monitoring the electromyographic feedback signal and gait changes after applying the stimulation scheme, updating the parameters of the stimulation scheme by calculating the signal change rate and motion trajectory error analysis, and obtaining an updated scheme, including: Calculating the root mean square and median frequency of the electromyographic feedback signal, and determining the muscle state according to the root mean square and median frequency; Obtaining a ground contact impact rate according to the gait change, and calculating a cushioning effectiveness index according to the ground contact impact rate; Constructing a lower limb motion trajectory according to the gait change, and calculating a mean square error between a joint angle and a standard gait according to the lower limb motion trajectory; The parameters in the stimulation scheme are adjusted according to the muscle state, the buffering effectiveness index and the mean square error to obtain an updated scheme.

10. A muscle stimulation and regulation system based on gait detection, characterized in that: include: A gait detection module, used to collect biomechanical data of the subject's lower limbs in real time through a multi-source sensor group of a wearable device; A gait recognition module is used to perform real-time phase recognition on the lower limb biomechanical data according to a dynamic programming algorithm, and construct a pressure center trajectory to extract features from the recognition results to obtain gait features; A gait analysis module is used to perform feature analysis on the gait characteristics through the constructed inverse dynamics model and the machine learning model, and output and determine the analysis results according to the dual-model collaborative decision-making mechanism to obtain a stimulation parameter decision result; A stimulation scheme generating module, used for generating an electrical stimulation waveform with phase synchronization characteristics according to the stimulation parameter decision result, and applying differentiated electrical stimulation to the target muscle during the gait cycle through the electrical stimulation waveform to obtain a stimulation scheme; The feedback adjustment module is used to monitor the myoelectric feedback signal and gait changes after the stimulation scheme is applied, and to update the parameters of the stimulation scheme by calculating the signal change rate and motion trajectory error analysis to obtain an updated scheme.

Images