A multi-mode fusion feedback-based foot drop electrical stimulation system and method

Abstract

A foot drop electrical stimulation system based on multi-mode fusion feedback, comprising an inertial measurement unit, an electromyographic information acquisition module, a near-infrared muscle oxygen acquisition module, an electrical stimulation intensity adjustment module, an electrical stimulation frequency adjustment module and a functional electrical stimulation module, the inertial measurement unit and the electromyographic information acquisition module are connected with the input side of the electrical stimulation intensity adjustment module, the output side of the electrical stimulation intensity adjustment module is connected with the input side of the functional electrical stimulation module, the near-infrared muscle oxygen acquisition module is connected with the electrical stimulation frequency adjustment module, the electrical stimulation frequency adjustment module is connected with the input side of the functional electrical stimulation module, and the output side of the functional electrical stimulation module is a working end for stimulating patients. And a foot drop electrical stimulation method based on multi-mode fusion feedback is provided. The present application improves the matching effect of nerve-muscle-limb multi-source information feedback and functional electrical stimulation assisted rehabilitation training, and realizes better foot drop rehabilitation training feedback effect.

Description

Technical Field

[0001] The present invention belongs to the field of limb movement rehabilitation training, and in particular relates to a foot drop electrical stimulation system and method based on multi-modal fusion feedback. Background Art

[0002] Foot drop, a common ankle dysfunction after stroke, is primarily manifested by paralysis or marked weakness of the ankle dorsiflexor muscles. This symptom typically stems from dysfunction of the common peroneal nerve, which controls multiple muscles, including the tibialis anterior, extensor digitorum longus, extensor digitorum brevis, extensor hallucis longus, and peroneus. Typical symptoms of foot drop include an inability to achieve the necessary dorsiflexion during the swing phase of gait, resulting in an inability to maintain adequate distance from the ground. Specifically, patients experience foot dragging on the affected side while walking. To compensate, they may take an abnormally high stepping motion or exhibit foot clearance problems and a circular gait in the lower limb. If left untreated for a long time, foot drop can lead to permanent structural changes in the tissues surrounding the ankle joint, causing joint contracture and deformity. Therefore, timely treatment and rehabilitation training to improve gait and enhance quality of life are crucial for patients and their families.

[0003] Functional electrical stimulation (FES) is an advanced rehabilitation technology used to assist in the treatment of post-stroke foot drop. This technology activates targeted muscles by applying low-frequency pulsed current, mimicking the natural movement patterns of healthy individuals. FES offers significant advantages in rehabilitation. It not only enhances muscle strength, promotes brain reorganization, strengthens neural impulses, and prevents muscle atrophy, but also improves overall muscle health. FES devices, with their compact size, ease of operation, portability, and low cost, significantly reduce the workload of rehabilitation physicians. However, FES technology has also encountered challenges in its practical application. Individual physiological differences complicate the selection of stimulation parameters, often relying on physician experience and patient feedback. This can result in over- or understimulation, making it difficult to meet individualized treatment needs. Furthermore, traditional FES devices lack the ability to monitor muscle status in real time. Continuous stimulation can cause muscle fatigue, and continued stimulation during this fatigued state can lead to damage. Furthermore, these devices typically utilize open-loop control schemes, and the stimulation waveform is often fixed and fixed. FES devices or systems also have some practical drawbacks, such as inconvenient operation, the inability to observe changes in muscle state in real time, and the inability to adjust FES parameters in real time to accommodate different medical needs. Although current research on feedback control methods for foot drop electrical stimulation has alleviated these issues to some extent, there is still room for improvement in integrating muscle state monitoring and muscle autonomy. Although some functional electrical stimulation devices incorporate feedback signals such as joint angle, acceleration, and plantar pressure, these physical signals cannot provide direct feedback on muscle state, ignoring muscle information and easily causing muscle fatigue. Some studies have attempted to address the issues of open-loop stimulation and lack of muscle state feedback by collecting surface electromyography (EMG) signals. However, focusing solely on surface EMG feedback leads to low accuracy in muscle state assessment and limited robustness of closed-loop control systems. Furthermore, most control methods fail to effectively integrate the rehabilitation subject's voluntary muscle strength and the detection of muscle fatigue during training, which are issues that need to be addressed in future research. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, the present invention provides a foot drop electrical stimulation system and method based on multimodal fusion feedback. This system combines multimodal muscle state information detection and its response mechanism, strengthening the coupling effect of each feedback link and improving the matching effect between nerve-muscle-limb multi-source information feedback and functional electrical stimulation-assisted rehabilitation training, thereby achieving better feedback effects for foot drop rehabilitation training. This innovative method is expected to provide more personalized and effective rehabilitation treatment plans for patients with foot drop.

[0005] The technical solution adopted by the present invention to solve its technical problem is:

[0006] A foot drop electrical stimulation system based on multimodal fusion feedback includes an inertial measurement unit, an electromyographic information acquisition module, a near-infrared muscle oxygen acquisition module, an electrical stimulation intensity adjustment module, an electrical stimulation frequency adjustment module, and a functional electrical stimulation module. The inertial measurement unit and the electromyographic information acquisition module are connected to the input side of the electrical stimulation intensity adjustment module, the output side of the electrical stimulation intensity adjustment module is connected to the input side of the functional electrical stimulation module, the near-infrared muscle oxygen acquisition module is connected to the electrical stimulation frequency adjustment module, and the electrical stimulation frequency adjustment module is connected to the input side of the functional electrical stimulation module. The output side of the functional electrical stimulation module is a working end for stimulating a patient.

[0007] Furthermore, the electrical stimulation intensity adjustment module includes a pre-training model, an electrical stimulation intensity adjustment module and a controller. In the pre-training model, the expected angle curve of the affected ankle joint will be used as the input of the electrical stimulation intensity model based on the affected ankle joint angle, the muscle activation state model based on the affected ankle joint angle, and the electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint, respectively obtaining the expected electrical stimulation intensity I E , desired muscle activation state and voluntary muscle strength on the affected side - electrical stimulation intensity I H The muscle activation state model is used to calculate and analyze the feedback surface electromyographic signal to obtain the actual muscle activation state of the affected side; the electrical stimulation intensity model based on the muscle activation state of the affected side will be used for muscle activation state feedback to obtain the corrected electrical stimulation intensity I A The electrical stimulation intensity adjustment module calculates and analyzes the expected electrical stimulation intensity, the corrected electrical stimulation intensity, and the autonomous muscle force-electrical stimulation intensity to obtain the required electrical stimulation intensity value I F The controller uses the control algorithm to fine-tune the required value of the electrical stimulation intensity and obtain the output value of the electrical stimulation intensity I O .

[0008] Furthermore, the foot drop electrical stimulation system further includes a system initialization unit and a model pre-training unit. The system initialization unit is connected to the model pre-training unit, the input side of the model pre-training unit is connected to the inertial measurement unit and the electromyography information acquisition module, and the output side of the model pre-training unit is connected to the input side of the electrical stimulation intensity adjustment module;

[0009] The system initialization unit includes wearing a foot drop electrical stimulation rehabilitation system based on multi-modal fusion feedback, initializing a functional electrical stimulation module, and calibrating various signal acquisition devices;

[0010] The model pre-training unit includes two parts: pre-experiment and model establishment. After the pre-experiment, the electrical stimulation intensity, acceleration, angular velocity and surface electromyography signals collected in the pre-experiment are used to establish a patient-personalized pre-training model, including: using the healthy-affected side mapping relationship to determine the expected angle curve of the affected ankle joint; using a nonlinear fitting method to establish an electrical stimulation intensity model based on the affected ankle joint angle; using an LSTM neural network model to establish an electrical stimulation intensity model based on the affected muscle activation state; using the relationship between the models to determine the establishment of a muscle activation state model based on the affected ankle joint angle and an electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint. After obtaining the above-mentioned pre-training models, these pre-training models are output to the electrical stimulation intensity adjustment module for storage for subsequent feedback control of the electrical stimulation intensity.

[0011] Furthermore, the inertial measurement unit is used to collect acceleration and angular velocity signals of the patient's healthy and affected calves and soles during foot drop rehabilitation training, calculate and analyze the current ankle joint angle, and then perform feedback control on the desired angle curve of the affected ankle joint;

[0012] The electromyographic information acquisition module is used for surface electromyographic signal acquisition, which is used to collect the surface electromyographic signals of the tibialis anterior muscle of the patient during foot drop rehabilitation training, calculate and analyze the current muscle activation state, and then provide feedback to adjust the electrical stimulation intensity;

[0013] The near-infrared muscle oxygen acquisition module uses a near-infrared light source and detector with wavelengths of 735nm and 850nm to collect near-infrared spectral signals of the tibialis anterior muscle of the patient during foot drop rehabilitation training, and uses the signals to feedback adjust the electrical stimulation frequency.

[0014] The electrical stimulation frequency adjustment module includes a muscle oxygen concentration change model, a muscle fatigue degree model and an electrical stimulation frequency adjustment model, wherein the muscle oxygen concentration change model is used to calculate and analyze the near-infrared spectral signal collected by the near-infrared muscle oxygen acquisition module to obtain a muscle oxygen concentration change value; the muscle fatigue degree model uses the obtained muscle oxygen concentration change to further calculate the muscle fatigue degree; and the electrical stimulation frequency adjustment model adjusts the electrical stimulation frequency parameters according to the muscle fatigue degree to obtain the electrical stimulation frequency demand value.

[0015] The functional electrical stimulation module is used to convert the electrical stimulation parameters sent by the electrical stimulation intensity adjustment module and the electrical stimulation frequency adjustment module into control instructions and execute them to output electrical stimulation pulses.

[0016] A foot drop electrical stimulation control method based on multi-mode fusion feedback comprises the following steps:

[0017] Step 1: System initialization, including wearing the foot drop electrical stimulation control system based on multimodal fusion feedback, initializing the functional electrical stimulation module, and calibrating each signal acquisition device;

[0018] Step 2: Model pre-training: conducting a pre-experiment on the patient to generate a pre-trained model, and saving the obtained pre-trained model in the electrical stimulation intensity adjustment module;

[0019] Step 3, using the expected angle of the affected ankle joint during rehabilitation training obtained from the pre-training model as an input of the control method;

[0020] Step 4: Use the actual ankle joint angle of the affected side at the current moment as angle feedback to adjust the expected ankle joint angle of the affected side in step 3;

[0021] Step 5: Input the expected angle of the affected ankle joint after feedback in step 4 into the electrical stimulation intensity model based on the affected ankle joint angle, the muscle activation state model based on the affected ankle joint angle, and the electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint to obtain the expected electrical stimulation intensity I E , desired muscle activation state and voluntary muscle strength on the affected side - electrical stimulation intensity I H ;

[0022] Step 6: Use the actual muscle activation state of the affected side at the current moment as the muscle activation state feedback to adjust the expected muscle activation state of the affected side in step 5, and input the expected muscle activation state of the affected side after feedback into the electrical stimulation intensity model based on the muscle activation state of the affected side to obtain the corrected electrical stimulation intensity I A ;

[0023] Step 7: Input the desired electrical stimulation intensity, voluntary muscle strength-electrical stimulation intensity and corrected electrical stimulation intensity obtained in steps 5 and 6 into the electrical stimulation intensity adjustment model to obtain the electrical stimulation intensity requirement value I F , and then the electrical stimulation intensity output value I is obtained through the control algorithm O At the same time, the near-infrared spectrum signal of the tibialis anterior muscle is collected and input into the electrical stimulation frequency adjustment module to calculate the changes in muscle oxygen concentration and muscle fatigue level. The muscle fatigue level is then fed back into the electrical stimulation frequency adjustment model to obtain the required electrical stimulation frequency value.

[0024] Step 8: Send the electrical stimulation intensity output value and the electrical stimulation frequency requirement value obtained in step 7 to the functional electrical stimulation module for control instruction conversion;

[0025] Step 9: The functional electrical stimulation module executes the control instruction to output a stimulation pulse to stimulate the peroneal nerve on the affected side of the patient.

[0026] Furthermore, in step 5, the electrical stimulation intensity model based on the ankle angle of the affected side is used to find the relationship between the patient's personalized ankle joint angle and the electrical stimulation intensity, which is established as follows:

[0027] 5.1.1. Obtain multiple sets of electrical stimulation intensity data and multiple sets of affected ankle joint angles, preprocess multiple sets of inertial measurement unit acceleration and angular velocity signals, and obtain multiple sets of affected ankle joint angles;

[0028] 5.1.2. Establish an electrical stimulation intensity model based on the angle of the affected ankle joint. Use a polynomial regression model to perform nonlinear fitting on multiple sets of data on the angle of the affected ankle joint and the intensity of the electrical stimulation. A second-order polynomial regression model is used for fitting. The model is expressed as:

[0029] I E =aθ H 2 +bθ H +c

[0030] Among them, I E is the desired electrical stimulation intensity; θ H is the ankle joint angle on the affected side; a, b, and c are the parameters of the model, which need to be determined by data fitting;

[0031] The muscle activation state model based on the ankle joint angle of the affected side is used to detect whether the muscle activation state of the tibialis anterior muscle meets the requirements during rehabilitation training, so as to correct the electrical stimulation intensity. It is established based on the electrical stimulation intensity model based on the muscle activation state of the affected side and the electrical stimulation intensity model based on the ankle joint angle of the affected side, and is expressed as:

[0032] M LSTM (W,H)=aθ H 2 +bθ H +c

[0033] Among them, M LSTM It is a model of electrical stimulation intensity based on the activation state of the affected muscle.

[0034] The electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint is used to characterize the magnitude of the patient's voluntary muscle strength of the tibialis anterior muscle at each moment corresponding to the desired angle curve of the affected ankle joint, so as to correct the electrical stimulation intensity. It is established as follows:

[0035] 5.2.1. Obtain the ankle joint angle curve and surface electromyographic signal (SEM) of the affected side during normal walking. Preprocess the acceleration and angular velocity data recorded by the two inertial measurement units on the affected side, as well as the SEM signal of the tibialis anterior muscle on the affected side, to obtain the ankle joint angle curve and SEM signal of the tibialis anterior muscle on the affected side during normal walking.

[0036] 5.2.2. Calculate the muscle activation state based on the muscle activation state model. Calculate the muscle activation state of the tibialis anterior muscle during normal walking using the same method as when establishing the electrical stimulation intensity model based on the muscle activation state of the affected side.

[0037] 5.2.3. Calculate the voluntary muscle force-electrical stimulation intensity. Use the muscle activation state calculated in 5.2.2 as the input to the electrical stimulation intensity model based on the muscle activation state of the affected side. Calculate the voluntary muscle force-electrical stimulation intensity corresponding to each desired ankle angle on the affected side.

[0038] 5.2.4. Based on the voluntary muscle force-electrical stimulation intensity corresponding to each desired angle of the affected ankle joint obtained in 5.2.3, a correction model of the electric stimulation intensity based on the voluntary muscle force of the affected ankle joint is obtained.

[0039] Furthermore, in step 6, the electrical stimulation intensity model based on the muscle activation state of the affected side is used to establish an electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint and a muscle activation state model based on the angle of the affected ankle joint, which is established as follows:

[0040] 6.1. Acquire multiple sets of electrical stimulation intensity data and multiple sets of affected-side surface electromyography signals, perform data preprocessing on the multiple sets of surface electromyography signals, and obtain the surface electromyography signal of the affected-side tibialis anterior muscle;

[0041] 6.2. Calculate the muscle activation state based on the muscle activation state model. Perform muscle synergy analysis on multiple sets of preprocessed surface electromyographic signals. Use the non-negative matrix factorization method to decompose them into a muscle activation weight matrix and a muscle activation sequence matrix. The muscle activation state model is as follows:

[0042] M≈W×H=M′,M∈R n×t ,W∈R n×k ,H∈R k×t

[0043] min||MM′|| 2 =∑(MM′) 2

[0044]

[0045] Where M is the preprocessed surface electromyography signal, W is the muscle activation weight matrix, H is the muscle activation sequence matrix, M′ is the matrix reconstructed based on the muscle activation weight matrix and the activation sequence matrix, R is a real number matrix, n is the number of collected muscles, t is the number of surface electromyography sampling points, k is the number of collaborative elements, and W T is the transposed matrix of the muscle activation weight matrix, H T is the transposed matrix of the muscle activation sequence matrix;

[0046] 6.3. Establish an electrical stimulation intensity model based on the activation state of the affected muscle. Use the multiple muscle activation weight matrices and muscle activation sequences obtained in step 6.2 and the multiple sets of electrical stimulation intensity data to construct a dataset. Use the multiple muscle activation weight matrices and muscle activation sequences as inputs to the LSTM neural network model, and use the multiple sets of electrical stimulation intensity data as labels for the LSTM neural network model. Train the neural network model to obtain an electrical stimulation intensity model based on the activation state of the affected muscle:

[0047] I=M LSTM (W,H)

[0048] Furthermore, in step 7, the electrical stimulation intensity is adjusted in the following manner to obtain an electrical stimulation intensity output value, and the process is as follows:

[0049] 7.1.1, the desired electrical stimulation intensity I obtained in step 5 and step 6 E , Corrected electrical stimulation intensity I A and voluntary muscle strength—electrical stimulation intensity I H Input into the electrical stimulation intensity adjustment model to obtain the electrical stimulation intensity requirement value I F , the electrical stimulation intensity regulation model is expressed as:

[0050] I F =I E -I H +αI A

[0051] Among them, I F is the required value of electrical stimulation intensity, I H is the voluntary muscle strength-electrical stimulation intensity, I A is the correction of the electrical stimulation intensity, α is the proportional coefficient;

[0052] 7.1.2, after calculating the required value of electrical stimulation intensity I F Then, the control algorithm of the controller is used to fine-tune the required value of the electrical stimulation intensity to obtain the output value of the electrical stimulation intensity I O , the PID controller is expressed as:

[0053]

[0054] Among them, I O (k) is the output value of the electrical stimulation intensity, I O (k-1) is the electrical stimulation intensity output value at the previous moment, K p , K i and K d They are the proportional, integral, and differential gains of the PID controller, and e(k) is the deviation at the current moment, i.e., I O (k)-IF (k), e(k-1) is the deviation of the previous moment, I F (k) is the required value of electrical stimulation intensity at the current moment, and e(i) is the deviation at the historical moment;

[0055] In step 7, the electrical stimulation frequency adjustment module uses the near-infrared spectrum signal collected by the near-infrared muscle oxygen collection module to perform feedback adjustment to obtain the required electrical stimulation frequency value. The process is as follows:

[0056] 7.2.1. Acquire near-infrared spectral signals at 735 nm and 850 nm from the affected tibialis anterior muscle;

[0057] 7.2.2. Calculate the concentration changes of oxyhemoglobin and deoxyhemoglobin based on the muscle oxygen concentration change model. Use the near-infrared spectral signal obtained in 7.2.1 to calculate the concentration changes of oxyhemoglobin and deoxyhemoglobin based on the modified Lambert-Beer law. The calculation method of the muscle oxygen concentration change model is as follows:

[0058]

[0059] Where ΔA represents the optical density, V t V is the voltage value measured by the detector at a certain moment during rehabilitation training, which is proportional to the light intensity. dark is the voltage when all wavelength lamps are off, V b is the baseline voltage signal, B is the differential path factor at each wavelength, ε is the molar absorption coefficient at each wavelength, ΔC HHb is the change in deoxyhemoglobin concentration, is the change value of oxygenated hemoglobin concentration, L is the distance between the light source and the detector;

[0060] 7.2.3. Calculate the degree of muscle fatigue according to the muscle fatigue degree model. The muscle fatigue degree model is established as follows: Calculate the total hemoglobin change according to the change value of deoxyhemoglobin concentration and the change value of oxyhemoglobin concentration. and oxygen saturation changes The degree of muscle fatigue is calculated using the change in oxygen saturation, using a linear relationship: F ​​= βΔStO2, where F is the degree of muscle fatigue, β is the proportional coefficient, and β>0;

[0061] 7.2.4. The electrical stimulation frequency demand value is obtained based on the electrical stimulation frequency adjustment model. The electrical stimulation frequency adjustment model is established as follows: the calculated muscle fatigue level is compared with the set muscle fatigue level threshold. The comparison rules are as follows: if the current muscle fatigue level is greater than the set muscle fatigue level threshold 1, the electrical stimulation frequency demand value will be set to 0 and electrical stimulation will be stopped; if the current muscle fatigue level is greater than the set muscle fatigue level threshold 2 but less than the set muscle fatigue level threshold 1, it is necessary to reduce the electrical stimulation pulse frequency to alleviate the muscle fatigue. For convenience, a linear relationship can be used for adjustment: f a =f1-γ(F-F2), where f a It is the required value of the electrical stimulation frequency after feedback of the muscle fatigue level, f1 is the electrical stimulation frequency when the muscle shows a complete tetanic contraction signal when the electrical stimulation intensity and pulse width are fixed, F2 is the set muscle fatigue level threshold 2, γ is the proportional coefficient, γ>0; if the current muscle fatigue level is less than the set muscle fatigue level threshold 2, the required value of the electrical stimulation frequency will remain unchanged.

[0062] Preferably, in step 3, the expected angle curve of the affected ankle joint is used to plan the patient's personalized ankle joint rehabilitation training angle curve, which is established as follows:

[0063] 3.1, obtain the ankle joint angle curve of the healthy side during normal walking; preprocess the acceleration and angular velocity data recorded by the two inertial measurement units on the healthy side to obtain the ankle joint angle curve of the healthy side during normal walking;

[0064] 3.2. Obtain the expected angle curve of the affected ankle joint during rehabilitation training. According to the certain consistency of the ankle joint angles of both legs during human walking, the angle curve of the healthy ankle joint during normal walking is used as the expected angle curve of the affected ankle joint during rehabilitation training.

[0065] The technical concept of the present invention is: the present invention analyzes multimodal signals through mapping relationships, nonlinear fitting and neural networks to customize personalized rehabilitation training plans for patients. Unlike traditional methods that only rely on switching of limited stimulation modes, this system can adjust electrical stimulation parameters in real time through a multi-layer feedback mechanism, accurately stimulate the peroneal nerve, trigger related muscle contraction, and thus achieve dorsiflexion of the ankle joint. This innovative design aims to improve the matching efficiency between nerve-muscle-limb multi-source information feedback and functional electrical stimulation-assisted rehabilitation training, so as to achieve the best foot drop rehabilitation training effect. The device can be used in conjunction with rehabilitation devices such as exoskeletons.

[0066] The beneficial effects of the present invention are mainly manifested in:

[0067] (1) Utilizing multimodal fusion feedback composed of inertial measurement units, surface electromyography signals, and near-infrared spectroscopy signals, functional electrical stimulation parameters are adjusted in real time, effectively improving the synergy between multi-source information feedback from nerves, muscles, and limbs and functional electrical stimulation-assisted rehabilitation training, providing better rehabilitation effects and experience;

[0068] (2) By analyzing the patient's muscle activation state through muscle synergy, the functional electrical stimulation parameters can be adjusted more specifically to achieve the appropriate muscle activation state and achieve better rehabilitation effects;

[0069] (3) By obtaining the patient's voluntary muscle strength on the affected side during autonomous walking, an electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint is established, fully considering the patient's voluntary muscle strength recovery during rehabilitation training, and improving the patient's autonomy in rehabilitation training and the efficiency of functional electrical stimulation rehabilitation;

[0070] (4) By using this system, the patient's expected ankle angle curve on the affected side, the electrical stimulation intensity model based on the angle of the affected ankle joint, the muscle activation state model based on the angle of the affected ankle joint, and the electrical stimulation intensity model based on the muscle activation state of the affected side can be obtained. This system can provide patients with personalized and adaptive rehabilitation training strategies and improve the effect of functional stimulation rehabilitation in patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 Schematic diagram of the structure of the foot drop electrical stimulation system based on multi-mode fusion feedback of the present invention;

[0072] Figure 2 This is a diagram showing the placement of the foot drop electrical stimulation system based on multi-modal fusion feedback on the lower limbs of a human body;

[0073] Figure 3 Flowchart for pre-training of the model of the present invention;

[0074] Figure 4 This is a flow chart of the electrical stimulation frequency adjustment module of the present invention;

[0075] Figure 5 This is a block diagram of the foot drop electrical stimulation control method based on multi-mode fusion feedback of the present invention. DETAILED DESCRIPTION

[0076] The present invention will be further described below with reference to the accompanying drawings.

[0077] Example 1

[0078] Reference Figures 1 to 4A foot drop electrical stimulation system based on multimodal fusion feedback includes a system initialization unit, a model pre-training unit, an inertial measurement unit, an electromyographic information acquisition module, a near-infrared muscle oxygen acquisition module, an electrical stimulation intensity adjustment module, an electrical stimulation frequency adjustment module, and a functional electrical stimulation module, wherein:

[0079] The system initialization unit includes wearing a foot drop electrical stimulation rehabilitation system based on multi-modal fusion feedback, initializing the functional electrical stimulation module and calibrating each signal acquisition device; a model pre-training unit, which is used to determine the expected angle curve of the affected ankle joint suitable for the patient's personalized rehabilitation training, the electrical stimulation intensity model based on the affected muscle activation state, the electrical stimulation intensity model based on the affected ankle joint angle, the muscle activation state model based on the affected ankle joint angle and the electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint, and output these pre-trained models to the electrical stimulation intensity adjustment module for storage for subsequent feedback control of the electrical stimulation intensity; an inertial measurement unit, which is used to collect the acceleration and angular velocity signals of the healthy and affected calves and soles of the feet of the patient during the foot drop rehabilitation training, and to perform feedback control on the electrical stimulation intensity; a surface electromyography signal acquisition module, which is used to collect the patient's The surface electromyographic signal of the tibialis anterior muscle during foot drop rehabilitation training is collected and used for feedback control of the electrical stimulation intensity; the near-infrared muscle oxygen acquisition module uses a near-infrared light source and detector with wavelengths of 735nm and 850nm to collect the near-infrared spectral signal of the tibialis anterior muscle during the patient's foot drop rehabilitation training, and uses the signal to feedback control the electrical stimulation frequency; the electrical stimulation intensity adjustment module is used to perform feedback calculation and analysis on the information output by the pre-training model, adjust the electrical stimulation intensity parameters, and use the control algorithm to fine-tune the electrical stimulation parameters; the electrical stimulation frequency adjustment module is used to calculate and analyze the near-infrared spectral signal collected by the near-infrared muscle oxygen acquisition module, obtain the degree of muscle fatigue, and adjust the electrical stimulation frequency parameters; the functional electrical stimulation module is used to convert the electrical stimulation parameters sent by the electrical stimulation intensity adjustment module and the electrical stimulation frequency adjustment module into control instructions and execute them, and output electrical stimulation pulses.

[0080] In this embodiment, the system initialization unit is performed as follows:

[0081] 1) Wear a foot drop electrical stimulation rehabilitation system based on multi-modal fusion feedback. Figure 2The figure shows the position of the foot drop electrical stimulation system based on multi-mode fusion feedback in the lower limbs of the human body according to the present invention. 1 represents the electrodes of the functional electrical stimulation module, 2 represents the electrodes of the electromyographic information acquisition module, 3 represents the near-infrared muscle oxygen acquisition module, and 4 represents the inertial measurement unit. The functional electrical stimulation module adopts a bipolar electrode configuration, and the two electrodes are placed at the body surface projection position below the head of the fibula, where the common peroneal nerve is closer to the skin surface, which is convenient for positioning and stimulation to induce the dorsiflexion movement of the ankle joint; the two electrodes of the surface electromyographic signal acquisition device are placed on the tibialis anterior muscle on the affected side; the near-infrared muscle oxygen acquisition device is placed on the tibialis anterior muscle on the affected side; the inertial measurement units are placed on the healthy and affected sides at 10 cm away from the ankle joint and on the dorsum of the foot, respectively, with a total of 2 groups of 4, and the inertial measurement unit-human posture alignment calibration is performed to convert the inertial measurement unit coordinate system into the human coordinate system;

[0082] 2) Initialize the functional electrical stimulation module. Initialize the electrical stimulation parameters, including initializing the electrical stimulation intensity, initializing the electrical stimulation frequency, and setting the electrical stimulation pulse width;

[0083] 3) Calibrate each signal acquisition device. This includes collecting baseline voltage signals and dark voltage data from the near-infrared muscle oxygen acquisition module; performing inertial measurement unit (IMU)-human body posture alignment calibration; and converting the IMU coordinate system to the human body coordinate system.

[0084] Model pre-training mainly includes two parts: after the system is initialized, two pre-experiments are carried out; based on the surface electromyography signals, near-infrared signals, acceleration and angular velocity collected in the two pre-experiments, the expected angle curve of the affected ankle joint suitable for the patient's personalized rehabilitation training, the electrical stimulation intensity model based on the affected muscle activation state, the electrical stimulation intensity model based on the affected ankle joint angle, the muscle activation state model based on the affected ankle joint angle, and the electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint are established in advance. After obtaining the above pre-training models, these pre-training models are output to the electrical stimulation intensity adjustment module for storage for subsequent feedback control of the electrical stimulation intensity.

[0085] like Figure 3The figure shows a flow chart of the pre-training of the model of the present invention. After the system is initialized, two pre-experiments are required on the patient to determine the pre-training model suitable for the patient. Pre-experiment ① is carried out in the following manner: the patient wears the foot drop electrical stimulation system based on multimodal fusion feedback of the present invention, and also wears an inertial measurement unit in the same position as the affected side on the healthy side; walks normally without the assistance of electrical stimulation; records the original signals of acceleration and angular velocity of the inertial measurement unit on the healthy and affected sides of the patient and the surface electromyography signal of the tibialis anterior muscle on the affected side during natural walking. Pre-experiment ② is carried out in the following manner: the patient wears the foot drop electrical stimulation system based on multimodal fusion feedback of the present invention; uses a functional electrical stimulation module to stimulate the peroneal nerve on the affected side of the patient, so that the patient's ankle joint can achieve dorsiflexion movement, fixes the electrical stimulation frequency and electrical stimulation pulse width during the period, and gradually increases the electrical stimulation intensity; synchronously records multiple sets of electrical stimulation intensity data, multiple sets of inertial measurement unit acceleration and angular velocity signals, and multiple sets of surface electromyography signals during the experiment.

[0086] Data preprocessing involves performing Kalman fusion on the acceleration and angular velocity signals from the inertial measurement unit (IMU) and the surface electromyography (SEM) signals. This involves performing Kalman fusion on the acceleration and angular velocity data recorded by each IMU to estimate the sensor attitude. The bias matrix is ​​then analyzed to determine the ankle joint angle by analyzing the attitude quaternions of the two IMUs. The SEM signals are then subjected to low-pass, high-pass, and notch filtering using a Butterworth filter to remove noise, skin motion artifacts, and power frequency interference.

[0087] The pre-trained model, derived from model pre-training, includes the desired angle curve for the affected ankle joint, an electrical stimulation intensity model based on the affected muscle activation state, an electrical stimulation intensity model based on the affected ankle joint angle, a muscle activation state model based on the affected ankle joint angle, and an electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint. These models are then output to the electrical stimulation intensity adjustment module and stored there for subsequent electrical stimulation intensity feedback control.

[0088] The information collected by the inertial measurement unit, the electromyographic information acquisition module and the near-infrared muscle oxygen acquisition module will be input into the electrical stimulation intensity adjustment module and the electrical stimulation frequency adjustment module as feedback signals to perform feedback adjustment of the electrical stimulation parameters.

[0089] The electrical stimulation intensity adjustment module includes a pre-training model, an electrical stimulation intensity adjustment model and a controller. In the pre-training model, the expected angle curve of the affected ankle joint will be used as the input of the electrical stimulation intensity model based on the affected ankle joint angle, the muscle activation state model based on the affected ankle joint angle and the electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint, to obtain the expected electrical stimulation intensity, the expected affected muscle activation state and the autonomous muscle strength-electrical stimulation intensity respectively; the muscle activation state model is used to calculate and analyze the feedback surface electromyography signal to obtain the actual muscle activation state of the affected side; the electrical stimulation intensity model based on the affected muscle activation state will be used for muscle activation state feedback to obtain the corrected electrical stimulation intensity; the electrical stimulation intensity adjustment model will calculate and analyze the expected electrical stimulation intensity, the corrected electrical stimulation intensity and the autonomous muscle strength-electrical stimulation intensity to obtain the required electrical stimulation intensity value; the controller uses the control algorithm to finely control the required electrical stimulation intensity value to obtain the output electrical stimulation intensity value.

[0090] The electrical stimulation frequency adjustment module includes a muscle oxygen concentration change model, a muscle fatigue degree model, and an electrical stimulation frequency adjustment model. The muscle oxygen concentration change model calculates and analyzes the near-infrared spectral signals collected by the near-infrared muscle oxygen acquisition module to determine muscle oxygen concentration changes. The muscle fatigue degree model uses the calculated muscle oxygen concentration changes to calculate the degree of muscle fatigue. The electrical stimulation frequency adjustment model determines the required electrical stimulation frequency based on the degree of muscle fatigue.

[0091] The functional electrical stimulation module is used to convert the electrical stimulation parameters sent by the electrical stimulation intensity adjustment module and the electrical stimulation frequency adjustment module into control instructions and execute them, output electrical stimulation pulses to the electrical stimulation electrodes, act on the patient's peroneal nerve, and induce dorsiflexion of the patient's affected side of the foot to achieve the target ankle joint angle.

[0092] Example 2

[0093] Reference Figure 5 , which is a block diagram of the foot drop electrical stimulation method based on multimodal fusion feedback of the present invention. The foot drop electrical stimulation method based on multimodal fusion feedback has been summarized, including the following steps:

[0094] Step 1: System initialization, including wearing the foot drop electrical stimulation rehabilitation system based on multimodal fusion feedback, initializing the functional electrical stimulation module, and calibrating each signal acquisition device;

[0095] Step 2: Model pre-training: conducting a pre-experiment on the patient to generate a pre-trained model, and saving the obtained pre-trained model in the electrical stimulation intensity adjustment module;

[0096] Step 3, using the expected angle of the affected ankle joint during rehabilitation training obtained from the pre-training model as an input of the control method;

[0097] The desired angle curve of the affected ankle joint is used to plan the patient's personalized ankle joint rehabilitation training angle curve, which is established as follows:

[0098] 3.1. Obtain the ankle joint angle curve of the healthy side during normal walking. Pre-process the acceleration and angular velocity data recorded by the two inertial measurement units on the healthy side in pre-experiment ① to obtain the ankle joint angle curve of the healthy side during normal walking.

[0099] 3.2. Obtain the expected angle curve of the affected ankle joint during rehabilitation training. The ankle joint angles of both legs show a certain consistency during walking. The normal walking ankle joint angle curve of the healthy side is used as the expected angle curve of the affected ankle joint during rehabilitation training.

[0100] Step 4: Use the actual ankle joint angle of the affected side at the current moment as angle feedback to adjust the expected ankle joint angle of the affected side in step 3;

[0101] Step 5: Input the desired angle of the affected ankle joint after feedback in step 4 into the electrical stimulation intensity model based on the angle of the affected ankle joint, the muscle activation state model based on the angle of the affected ankle joint, and the electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint, to obtain the desired electrical stimulation intensity, the desired voluntary muscle activation state, and the electrical stimulation intensity.

[0102] The electrical stimulation intensity model based on the ankle angle of the affected side is used to find the relationship between the patient's personalized ankle angle and electrical stimulation intensity. It is established as follows:

[0103] 5.1.1. Obtain multiple sets of electrical stimulation intensity data and multiple sets of affected ankle joint angles. Preprocess the multiple sets of inertial measurement unit acceleration and angular velocity signals from pre-experiment ② to obtain multiple sets of affected ankle joint angles.

[0104] 5.1.2. Establish an electrical stimulation intensity model based on the angle of the affected ankle joint. Use a polynomial regression model to perform nonlinear fitting on multiple sets of data on the angle of the affected ankle joint and multiple sets of electrical stimulation intensity. For convenience, a second-order polynomial regression model is used for fitting. The model is expressed as:

[0105] I E =aθ H 2 +bθ H +c

[0106] Among them, I E is the desired electrical stimulation intensity; θ H is the ankle joint angle of the affected side; a, b and c are the parameters of the model, which need to be determined through data fitting.

[0107] The muscle activation state model based on the ankle joint angle on the affected side is used to detect whether the muscle activation state of the tibialis anterior muscle meets the requirements during rehabilitation training, thereby correcting the electrical stimulation intensity. It is established based on the electrical stimulation intensity model based on the muscle activation state of the affected side and the electrical stimulation intensity model based on the ankle joint angle on the affected side, and is expressed as:

[0108] M LSTM (W,H)=aθ H 2 +bθ H +c

[0109] Among them, M LSTM It is a model of electrical stimulation intensity based on the activation state of the affected muscle.

[0110] The electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint is used to characterize the magnitude of the patient's voluntary muscle strength of the tibialis anterior muscle at each moment corresponding to the desired angle curve of the affected ankle joint, so as to correct the electrical stimulation intensity. It is established as follows:

[0111] 5.2.1. Obtain the ankle joint angle curve and surface electromyographic signal (SEM) of the affected side during normal walking. Preprocess the acceleration and angular velocity data recorded by the two inertial measurement units on the affected side in Preliminary Experiment ①, as well as the SEM signal of the tibialis anterior muscle on the affected side, to obtain the ankle joint angle curve and SEM signal of the tibialis anterior muscle on the affected side during normal walking.

[0112] 5.2.2. Calculate the muscle activation state based on the muscle activation state model. Calculate the muscle activation state of the tibialis anterior muscle during normal walking using the same method as when establishing the electrical stimulation intensity model based on the muscle activation state of the affected side.

[0113] 5.2.3. Calculate the voluntary muscle force-electrical stimulation intensity. The muscle activation state calculated in 5.2.2 is used as the input to the electrical stimulation intensity model based on the muscle activation state of the affected side. The voluntary muscle force-electrical stimulation intensity corresponding to each desired ankle angle on the affected side is obtained.

[0114] 5.2.4. Based on the voluntary muscle force-electrical stimulation intensity corresponding to each desired angle of the affected ankle joint obtained in 5.2.3, obtain a correction model for the electrical stimulation intensity based on the voluntary muscle force of the affected ankle joint;

[0115] Step 6: Using the actual muscle activation state of the affected side at the current moment as muscle activation state feedback, feedback adjustment is performed on the expected muscle activation state of the affected side in step 5, and the expected muscle activation state after feedback is input into the electrical stimulation intensity model based on the muscle activation state of the affected side to obtain a corrected electrical stimulation intensity;

[0116] The electrical stimulation intensity model based on the muscle activation state of the affected side is used to establish the electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint and the muscle activation state model based on the angle of the affected ankle joint. It is established as follows:

[0117] 6.1. Obtain multiple sets of electrical stimulation intensity data and multiple sets of surface electromyographic signals on the affected side. Perform data preprocessing on the multiple sets of surface electromyographic signals from pre-experiment ② to obtain the surface electromyographic signals of the tibialis anterior muscle on the affected side.

[0118] 6.2. Calculate the muscle activation state based on the muscle activation state model. Perform muscle synergy analysis on multiple sets of pre-processed surface electromyographic signals and decompose them into a muscle activation weight matrix and a muscle activation sequence matrix using the non-negative matrix factorization method. The muscle activation state model is as follows:

[0119] M≈W×H=M′,M∈R n×t ,W∈R n×k ,H∈R k×t

[0120] min||MM′|| 2 =∑(MM′) 2

[0121]

[0122] Where M is the preprocessed surface electromyography signal, W is the muscle activation weight matrix, H is the muscle activation sequence matrix, M′ is the matrix reconstructed based on the muscle activation weight matrix and the activation sequence matrix, R is a real number matrix, n is the number of collected muscles, t is the number of surface electromyography sampling points, k is the number of collaborative elements, and W T is the transposed matrix of the muscle activation weight matrix, H T is the transposed matrix of the muscle activation sequence matrix;

[0123] 6.3. Establish an electrical stimulation intensity model based on the activation state of the affected muscle. Use the multiple muscle activation weight matrices and muscle activation sequences obtained in step 6.2 and the multiple sets of electrical stimulation intensity data to construct a dataset. Use the multiple muscle activation weight matrices and muscle activation sequences as inputs to the LSTM neural network model, and the multiple sets of electrical stimulation intensity data as labels for the LSTM neural network model. Train the neural network model to obtain an electrical stimulation intensity model based on the activation state of the affected muscle:

[0124] I=M LSTM (W,H)

[0125] Among them, M LSTM It is a model of electrical stimulation intensity based on the activation state of the affected muscle;

[0126] Step 7: Input the desired electrical stimulation intensity, voluntary muscle strength-electrical stimulation intensity and corrected electrical stimulation intensity obtained in steps 5 and 6 into the electrical stimulation intensity adjustment model to obtain the electrical stimulation intensity requirement value I F , and then the controller calculates and analyzes the output value of the electrical stimulation intensity I O At the same time, the near-infrared spectrum signal of the tibialis anterior muscle is collected and input into the electrical stimulation frequency adjustment module to calculate the changes in muscle oxygen concentration and muscle fatigue. The muscle fatigue level is then fed back into the electrical stimulation frequency adjustment model to obtain the required electrical stimulation frequency value.

[0127] The electrical stimulation intensity adjustment module includes a pre-training model, an electrical stimulation intensity adjustment model and a controller. The electrical stimulation intensity is adjusted in the following way to obtain the electrical stimulation intensity output value I O , the process is as follows:

[0128] 7.1.1. Use the pre-trained model to perform steps 3 to 6 to obtain the desired electrical stimulation intensity I E , Corrected electrical stimulation intensity I A and voluntary muscle strength—electrical stimulation intensity I H

[0129] 7.1.2. Input the desired electrical stimulation intensity, corrected electrical stimulation intensity, and voluntary muscle strength-electrical stimulation intensity obtained in 7.1.1 into the electrical stimulation intensity adjustment model to obtain the required electrical stimulation intensity value. The electrical stimulation intensity adjustment model is expressed as:

[0130] I F =I E -I H +αI A

[0131] Among them, I F is the required value of electrical stimulation intensity, I H is the voluntary muscle strength-electrical stimulation intensity, I A is used to correct the electrical stimulation intensity, and α is the proportional coefficient.

[0132] 7.1.3. After calculating the required electrical stimulation intensity value, the controller's control algorithm is used to fine-tune the required electrical stimulation intensity value to obtain the electrical stimulation intensity output value. Take the PID control algorithm as an example to dynamically adjust the required electrical stimulation intensity value. PID control aims to achieve rapid response and reduce steady-state errors, enabling the system to more accurately track the required electrical stimulation intensity value while reducing errors caused by changes in muscle state or external interference, thereby ensuring the safety and effectiveness of electrical stimulation therapy. Specifically, the PID controller can be expressed as:

[0133]

[0134] Among them, IO (k) is the output value of the electrical stimulation intensity, I O (k-1) is the electrical stimulation intensity output value at the previous moment, K p , K i and K d They are the proportional, integral, and differential gains of the PID controller, and e(k) is the deviation at the current moment, i.e., I O (k)-I F (k), e(k-1) is the deviation of the previous moment, I F (k) is the required value of electrical stimulation intensity at the current moment, and e(i) is the deviation at the historical moment.

[0135] The electrical stimulation frequency adjustment module uses the near-infrared spectrum signal collected by the near-infrared muscle oxygen acquisition module for feedback adjustment to obtain the required electrical stimulation frequency value. Near-infrared spectroscopy technology can monitor the changes in the concentration of oxyhemoglobin and deoxyhemoglobin in muscles in real time. These changes directly reflect the oxidative metabolism state of the muscles and the degree of muscle fatigue, such as Figure 4 As shown in the flowchart of the electrical stimulation frequency adjustment module of the present invention, the electrical stimulation frequency requirement value is obtained by the following method:

[0136] 7.2.1. Acquire near-infrared spectral signals at 735 nm and 850 nm from the affected tibialis anterior muscle;

[0137] 7.2.2. Calculate the concentration changes of oxyhemoglobin and deoxyhemoglobin based on the muscle oxygen concentration change model. Use the near-infrared spectral signal obtained in 7.2.1 to calculate the concentration changes of oxyhemoglobin and deoxyhemoglobin based on the modified Lambert-Beer law. The calculation method of the muscle oxygen concentration change model is as follows:

[0138]

[0139] Where ΔA represents the optical density, V t V is the voltage value measured by the detector at a certain moment during rehabilitation training, which is proportional to the light intensity. dark is the voltage when all wavelength lamps are off, V b is the baseline voltage signal, B is the differential path factor at each wavelength, ε is the molar absorption coefficient at each wavelength, ΔC HHb is the change in deoxyhemoglobin concentration, is the change value of oxygenated hemoglobin concentration, L is the distance between the light source and the detector;

[0140] 7.2.3, calculate the degree of muscle fatigue according to the muscle fatigue degree model. The muscle fatigue degree model is established as follows: calculate the total hemoglobin change value according to the change value of deoxyhemoglobin concentration and the change value of oxyhemoglobin concentration and oxygen saturation changes The degree of muscle fatigue is calculated using changes in oxygen saturation. For convenience, a linear relationship can be used as an example for calculation: F = βΔStO2, where F is the degree of muscle fatigue, β is the proportional coefficient, and β>0.

[0141] 7.2.4. Obtain the required value of the electrical stimulation frequency according to the electrical stimulation frequency adjustment model. The electrical stimulation frequency adjustment model is established as follows: Compare the calculated muscle fatigue level with the set muscle fatigue threshold. The comparison rules are as follows: If the current muscle fatigue level is greater than the set muscle fatigue threshold 1, the required value of the electrical stimulation frequency will be set to 0, and electrical stimulation will be stopped; if the current muscle fatigue level is greater than the set muscle fatigue threshold 2 and less than the set muscle fatigue threshold 1, it is necessary to reduce the electrical stimulation pulse frequency to relieve the muscle fatigue. For convenience, a linear relationship can be used for adjustment: f a =f1-γ(F-F2), where f a It is the required value of the electrical stimulation frequency after feedback of the muscle fatigue level, f1 is the electrical stimulation frequency when the muscle shows a complete tetanic contraction signal when the electrical stimulation intensity and pulse width are fixed, F2 is the set muscle fatigue level threshold 2, γ is the proportional coefficient, γ>0; if the current muscle fatigue level is less than the set muscle fatigue level threshold 2, the required value of the electrical stimulation frequency will remain unchanged.

[0142] Step 8: Send the electrical stimulation intensity output value and the electrical stimulation frequency requirement value obtained in step 7 to the functional electrical stimulation module for control instruction conversion;

[0143] Step 9: The functional electrical stimulation module executes the control instruction to output a stimulation pulse to stimulate the peroneal nerve on the affected side of the patient.

[0144] The functional electrical stimulation module in this invention outputs electrical stimulation signals using low-frequency, symmetrically balanced biphasic pulses. The positive phase pulses provide electrical stimulation, while the negative phase pulses provide balance. This reduces the likelihood of pain or discomfort and delays muscle fatigue. Both the positive and negative phase pulses are current-type rectangular pulses, with an electrical stimulation intensity of 0-100 mA, a pulse width of 200-700 μs, and a pulse frequency of 0-100 Hz.

[0145] The embodiments of this specification are merely examples of implementations of the invention and are provided for illustrative purposes only. The scope of protection of the present invention should not be considered limited to the specific embodiments described in these embodiments. The scope of protection of the present invention also extends to equivalent technical means that can be conceived by a person of ordinary skill in the art based on the invention.

Claims

1. A foot drop electrical stimulation system based on multimodal fusion feedback, characterized in that the system includes an inertial measurement unit, an electromyographic information acquisition module, a near-infrared muscle oxygen acquisition module, an electrical stimulation intensity adjustment module, an electrical stimulation frequency adjustment module and a functional electrical stimulation module, wherein the inertial measurement unit and the electromyographic information acquisition module are connected to the input side of the electrical stimulation intensity adjustment module, the output side of the electrical stimulation intensity adjustment module is connected to the input side of the functional electrical stimulation module, the near-infrared muscle oxygen acquisition module is connected to the electrical stimulation frequency adjustment module, the electrical stimulation frequency adjustment module is connected to the input side of the functional electrical stimulation module, and the output side of the functional electrical stimulation module is a working end for stimulating a patient; The electrical stimulation intensity adjustment module includes a pre-training model, an electrical stimulation intensity adjustment model and a controller. In the pre-training model, the expected angle curve of the affected ankle joint will be used as the input of the electrical stimulation intensity model based on the affected ankle joint angle, the muscle activation state model based on the affected ankle joint angle and the electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint, and the expected electrical stimulation intensity is obtained respectively. I E , desired muscle activation state and voluntary muscle strength on the affected side - electrical stimulation intensity I H The muscle activation state model is used to calculate and analyze the feedback surface electromyographic signal to obtain the actual muscle activation state of the affected side; the electrical stimulation intensity model based on the muscle activation state of the affected side will be used for muscle activation state feedback to obtain the corrected electrical stimulation intensity. I A The electrical stimulation intensity regulation model calculates and analyzes the expected electrical stimulation intensity, the corrected electrical stimulation intensity, and the voluntary muscle force-electrical stimulation intensity to obtain the required electrical stimulation intensity value. I F The controller uses the control algorithm to fine-tune the required value of the electrical stimulation intensity and obtain the output value of the electrical stimulation intensity. I o ; The electrical stimulation intensity model based on the ankle angle of the affected side is used to find the relationship between the ankle angle and the electrical stimulation intensity that is personalized for the patient. It is established as follows: 1.

1. Obtain multiple sets of electrical stimulation intensity data and multiple sets of affected ankle joint angles through preliminary experiments. Preprocess multiple sets of inertial measurement unit acceleration and angular velocity signals to obtain multiple sets of affected ankle joint angles. 1.

2. Establish an electrical stimulation intensity model based on the angle of the affected ankle joint. Use a polynomial regression model to perform nonlinear fitting on multiple sets of data on the angle of the affected ankle joint and the intensity of the electrical stimulation. A second-order polynomial regression model is used for fitting. The model is expressed as: ； in, I E is the desired electrical stimulation intensity, is the ankle angle of the affected side; 、 b and c are the parameters of the model, which need to be determined by data fitting; The muscle activation state model based on the ankle joint angle of the affected side is used to detect whether the muscle activation state of the tibialis anterior muscle meets the requirements during rehabilitation training, so as to correct the electrical stimulation intensity. It is established based on the electrical stimulation intensity model based on the muscle activation state of the affected side and the electrical stimulation intensity model based on the ankle joint angle of the affected side, and is expressed as: ； in It is a model of electrical stimulation intensity based on the activation state of the affected muscle; The electrical stimulation intensity correction model based on the voluntary muscle strength of the affected ankle joint is used to characterize the magnitude of the patient's voluntary muscle strength of the tibialis anterior muscle at each moment corresponding to the desired angle curve of the affected ankle joint, so as to correct the electrical stimulation intensity. It is established as follows: 2.

1. Obtain the ankle joint angle curve and surface electromyography signal of the affected side during normal walking. Preprocess the acceleration and angular velocity data recorded by the two inertial measurement units on the affected side and the surface electromyography signal of the tibialis anterior muscle on the affected side to obtain the ankle joint angle curve and surface electromyography signal of the tibialis anterior muscle on the affected side during normal walking. 2.

2. Calculate the muscle activation state based on the muscle activation state model. Calculate the muscle activation state of the surface electromyographic signal of the tibialis anterior muscle during normal walking on the affected side using the same method as when establishing the electrical stimulation intensity model based on the muscle activation state of the affected side. 2.

3. Calculate the voluntary muscle force-electrical stimulation intensity. Use the muscle activation state calculated in 2.2 as the input of the electrical stimulation intensity model based on the muscle activation state of the affected side to obtain the voluntary muscle force-electrical stimulation intensity corresponding to each desired ankle angle on the affected side. 2.4, according to the voluntary muscle force-electrical stimulation intensity corresponding to each desired angle of the affected ankle joint obtained in 2.3, obtain an electrical stimulation intensity correction model based on the voluntary muscle force of the affected ankle joint; Desired electrical stimulation intensity I E , correct the intensity of electrical stimulation I A and voluntary muscle strength—electrical stimulation intensity I H Input into the electrical stimulation intensity adjustment model to obtain the electrical stimulation intensity requirement value I F , the electrical stimulation intensity regulation model is expressed as: ； in, I F is the required value of electrical stimulation intensity, I H is the voluntary muscle strength - electrical stimulation intensity, I A To correct the intensity of electrical stimulation, is the proportional coefficient.

2. A foot drop electrical stimulation system based on multimodal fusion feedback according to claim 1, characterized in that: It also includes a system initialization unit and a model pre-training unit, wherein the system initialization unit is connected to the model pre-training unit, the input side of the model pre-training unit is connected to the inertial measurement unit and the electromyographic information acquisition module, and the output side of the model pre-training unit is connected to the input side of the electrical stimulation intensity adjustment module; The system initialization unit includes wearing a foot drop electrical stimulation system based on multi-modal fusion feedback, initializing a functional electrical stimulation module, and calibrating various signal acquisition devices; The model pre-training unit includes two parts: a pre-experiment and a model establishment. After the pre-experiment, the electrical stimulation intensity, acceleration, angular velocity and surface electromyography signals collected in the pre-experiment are used to establish a patient-personalized pre-training model, including determining the expected angle curve of the affected ankle joint using the healthy-affected side mapping relationship; establishing an electrical stimulation intensity model based on the affected ankle joint angle using a nonlinear fitting method; establishing an electrical stimulation intensity model based on the affected muscle activation state using an LSTM neural network model; and determining the establishment of a muscle activation state model based on the affected ankle joint angle and an electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint using the mutual relationship between the models. After obtaining the above-mentioned pre-training models, these pre-training models are output to the electrical stimulation intensity adjustment module for storage for subsequent feedback control of the electrical stimulation intensity.

3. A foot drop electrical stimulation system based on multimodal fusion feedback according to claim 1 or 2, characterized in that: The inertial measurement unit is used to collect acceleration and angular velocity signals of the patient's healthy and affected calves and soles during foot drop rehabilitation training, calculate and analyze the current ankle joint angle, and then perform feedback control on the desired angle curve of the affected ankle joint; The electromyographic information acquisition module is used for surface electromyographic signal acquisition, which is used to collect the surface electromyographic signals of the tibialis anterior muscle of the patient during foot drop rehabilitation training, calculate and analyze the current muscle activation state, and then provide feedback to adjust the electrical stimulation intensity; The near-infrared muscle oxygen acquisition module uses a near-infrared light source and detector with wavelengths of 735nm and 850nm to collect near-infrared spectral signals of the tibialis anterior muscle of the patient during foot drop rehabilitation training, and uses the signals to feedback adjust the electrical stimulation frequency.

4. A foot drop electrical stimulation system based on multimodal fusion feedback according to claim 1 or 2, characterized in that: The electrical stimulation frequency adjustment module includes a muscle oxygen concentration change model, a muscle fatigue degree model and an electrical stimulation frequency adjustment model, wherein the muscle oxygen concentration change model is used to calculate and analyze the near-infrared spectral signal collected by the near-infrared muscle oxygen acquisition module to obtain a muscle oxygen concentration change value; the muscle fatigue degree model uses the obtained muscle oxygen concentration change to further calculate the muscle fatigue degree; and the electrical stimulation frequency adjustment model adjusts the electrical stimulation frequency parameters according to the muscle fatigue degree to obtain the electrical stimulation frequency demand value.

5. A foot drop electrical stimulation system based on multimodal fusion feedback according to claim 1 or 2, characterized in that: The electrical stimulation intensity model based on the muscle activation state of the affected side is used to establish an electrical stimulation intensity correction model based on the autonomous muscle strength of the affected ankle joint and a muscle activation state model based on the angle of the affected ankle joint, which is established as follows: 3.

1. Acquire multiple sets of electrical stimulation intensity data and multiple sets of affected-side surface electromyography signals, perform data preprocessing on the multiple sets of surface electromyography signals, and obtain the surface electromyography signals of the affected-side tibialis anterior muscle; 3.

2. Calculate the muscle activation state according to the muscle activation state model. Perform muscle synergy analysis on multiple sets of preprocessed surface electromyographic signals. Decompose them into muscle activation weight matrix and muscle activation sequence matrix using non-negative matrix factorization. The muscle activation state model is as follows: ； ； ； in, M is the preprocessed surface electromyography signal, W is the muscle activation weight matrix, H is the muscle activation sequence matrix, is the matrix reconstructed based on the muscle activation weight matrix and the activation sequence matrix, R is a real number matrix, n To collect muscle numbers, t is the number of surface electromyography sampling points, k is the number of collaborative elements, is the transposed matrix of the muscle activation weight matrix, is the transposed matrix of the muscle activation sequence matrix; 3.

3. Establish an electrical stimulation intensity model based on the activation state of the affected muscle. Use the multiple muscle activation weight matrices and muscle activation sequences obtained in step 3.2 and multiple sets of electrical stimulation intensity data to construct a dataset. Use the multiple muscle activation weight matrices and muscle activation sequences as inputs to the LSTM neural network model, and use the multiple sets of electrical stimulation intensity data as labels for the LSTM neural network model. Train the neural network model to obtain an electrical stimulation intensity model based on the activation state of the affected muscle: ； in, I is the electrical stimulation intensity output by the electrical stimulation intensity model based on the activation state of the affected muscle, It is a model of electrical stimulation intensity based on the activation state of the affected muscle.

Images