Mirror image dynamic balance training method and system for hemiplegic patient

Abstract

The invention discloses a mirror image dynamic balance training method and system for a hemiplegic patient, and relates to the technical field of motion intention recognition, and the method comprises the steps: collecting muscle contraction change data of the healthy side lower limb of the patient, and carrying out the preprocessing; feature data fusion is carried out on the preprocessed data, the data are input into a TGWOA-KNN model for classification, and the action intention of the patient is recognized; the affected side exoskeleton assists the affected side lower limb of the patient in mirroring the action of the uninjured side lower limb according to the action intention of the patient identified by the TGWOA-KNN model. According to the invention, the TGWOA-KNN model is used to improve the recognition effect on the limb movement of the subject; an FMG sensor and an IMU sensor are utilized, an intention recognition and posture mapping scheme is established, and dynamic balance training is completed in the mode that an exoskeleton assists an affected side lower limb in mirroring an uninjured side lower limb to move; meanwhile, visual stimulation is performed through visual display in training, so that the rehabilitation initiative of the patient is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motion intention recognition, and more specifically, to a mirror dynamic balance training method and system for hemiplegic patients. Background Art

[0002] In the early stage of hemiplegia, the core muscle groups of patients are weak, and it is difficult to maintain their own balance during movement. Moreover, when walking, they mainly rely on body swing compensation, resulting in uncoordinated movement gaits. Therefore, it is necessary to train the core muscle groups of patients and the symmetrical activities of the bilateral lower limb movement muscle groups through dynamic balance actions such as squatting and standing up. It is a relatively popular training method in current rehabilitation hospitals for rehabilitation therapists to use lower limb rehabilitation exoskeletons to assist hemiplegic patients in rehabilitation training.

[0003] This method belongs to an active rehabilitation training method, so the recognition of motion intention is required. Currently, the recognition of lower limb motion intention mostly uses electromyographic signals, but electromyographic sensors are easily affected by external interference, and situations such as muscle fatigue and sweating will also affect the accuracy. In contrast, the method of measuring the contraction changes of the motion muscle groups through a force myography (FMG) sensor to interpret the limb movement intention can more efficiently reflect the limb movement situation. The IMU sensor has the advantages of real-time monitoring of the object's posture and rapid response, so the IMU sensor and the FMG sensor can be used simultaneously to recognize the lower limb motion intention.

[0004] Therefore, how to complete dynamic balance training through a lower limb rehabilitation exoskeleton for the purpose of enhancing the core muscle groups of patients, promoting the coordinated movement of the bilateral lower limb movement muscle groups, and correcting the wrong compensation mode during the patient's movement, how to achieve the detection of the active intention of the lower limbs suitable for hemiplegic patients, and improve the recognition accuracy are problems that need to be urgently solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a mirror dynamic balance training method and system for hemiplegic patients, which use an exoskeleton to assist in enhancing the core muscle groups of patients and promoting the coordinated movement of the bilateral lower limb movement muscle groups, so as to correct the wrong compensation mode during the patient's movement, establish a healthy lower limb intention recognition and posture mapping scheme using FMG and IMU, and improve the accuracy during the recognition of motion intention. To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A mirror dynamic balance training method for hemiplegic patients includes:

[0007] Collecting the muscle contraction change data of the healthy lower limb of the patient;

[0008] Preprocessing the collected muscle contraction change data;

[0009] Perform feature data fusion on the preprocessed data, input it into the TGWOA-KNN model for classification, and identify the patient's action intention;

[0010] The affected-side exoskeleton assists the affected lower limb of the patient to mirror the actions of the healthy lower limb according to the patient's action intention recognized by the TGWOA-KNN model.

[0011] Optionally, the collection of muscle contraction change data of the patient's healthy lower limb includes: collecting muscle contraction change data of the patient's healthy lower limb through a wearable FMG sensor array; and collecting attitude data of the patient's healthy lower limb through an IMU sensor.

[0012] Optionally, the preprocessing of the collected muscle contraction change data includes: processing the collected FMG signal and IMU signal using a Butterworth low-pass filter, then extracting the FMG and IMU feature data therein, and normalizing the FMG feature data.

[0013] Optionally, the feature data fusion of the preprocessed data includes: fusing the two types of data using a feature-level fusion method.

[0014] Optionally, the establishment steps of the TGWOA-KNN model include: optimizing the K value of the KNN model using an improved WOA optimization algorithm; initializing the population using a Tent mapping strategy; and increasing the population diversity using a Gaussian mutation operator.

[0015] Optionally, the population initialization function generated by using the Tent mapping strategy to initialize the population is:

[0016]

[0017] In the formula, N is the current iteration number, t is the chaotic sequence at the current iteration number, d is the serial number of the chaotic mapping sequence, and X max is the upper boundary, and X min is the lower boundary;

[0018] The position update function after increasing the population diversity using the Gaussian mutation operator is:

[0019]

[0020] In the formula, X best is the optimal solution of the current algorithm, and G(0,1) is a Gaussian distribution random vector.

[0021] Optionally, after the feature data fusion of the preprocessed data, it further includes:

[0022] Using a complementary filter to fuse and calculate the angle:

[0023]

[0024] In the formula, θ filtered is the filtered angle, α is the weight of the filter, θ g is the gyroscope angle, ω is the angular velocity measured by the gyroscope, Δt is the time interval, θ a is the accelerometer angle, θ t-1 is the gyroscope angle at the previous moment, a1, a2, and a3 are the measured values of the accelerometer on the axes of the space rectangular coordinate system;

[0025] Obtain the healthy-side joint angle:

[0026]

[0027] In the formula, θ0 is the reference point, θ T is the thigh angle information, θ C is the calf angle information, θ H and θ K are the hip and knee joint angles, and are simultaneously mapped to the motion angles of the affected-side exoskeleton.

[0028] Optionally, it further includes: calculating the information between multi-point IMU signals to obtain the squatting and standing height:

[0029] The calculation formula for the height change of the IMU sensor in space is:

[0030] h = h t-1 + v cur ·Δt;

[0031] In the formula, h t-1 is the height of the IMU sensor at the previous moment, v cur is the current velocity obtained by integrating the accelerometer, Δt is the time interval, and h is the squatting and standing height of the current limb.

[0032] Optionally, a mirror dynamic balance training system for hemiplegic patients includes:

[0033] Acquisition module: used to acquire the muscle contraction change data of the healthy-side lower limb of the patient;

[0034] Preprocessing module: used to preprocess the acquired muscle contraction change data;

[0035] Recognition module: used to perform feature data fusion on the preprocessed data, input it into the TGWOA-KNN model for classification, and recognize the patient's action intention;

[0036] Execution module: used to control the affected-side exoskeleton to assist the affected-side lower limb of the patient to mirror the actions of the healthy-side lower limb according to the patient's action intention recognized by the TGWOA-KNN model.

[0037] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a mirror dynamic balance training method and system for hemiplegic patients, which has the following beneficial effects:

[0038] The present invention proposes a mirror dynamic balance training method for hemiplegic patients, including: collecting muscle contraction change data of the healthy lower limb of the patient; preprocessing the collected muscle contraction change data; performing feature data fusion on the preprocessed data, inputting it into the TGWOA-KNN model for classification to identify the patient's movement intention; the affected exoskeleton assists the affected lower limb of the patient to mirror the movement of the healthy lower limb according to the patient's movement intention recognized by the TGWOA-KNN model. The present invention proposes a TGWOA algorithm, preferentially selects the K value in the KNN algorithm, and then uses the TGWOA-KNN model to improve the classification effect of the subject's limb movements, providing more efficient intention perception for the rehabilitation exoskeleton robot; the present invention uses FMG sensors and IMU sensors to establish an intention recognition and attitude mapping scheme, and completes dynamic balance training by means of the exoskeleton assisting the affected lower limb to mirror the movement of the healthy lower limb, so as to achieve the purpose of strengthening the patient's core muscles and coordinating the recovery of bilateral lower limb movement muscles. At the same time, visual stimulation is carried out through visual display during training to improve the patient's rehabilitation initiative. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] 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 for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0040] Figure 1 It is a schematic flow chart of a mirror dynamic balance training method for hemiplegic patients provided by the present invention.

[0041] Figure 2 It is a training flow chart of the TGWOA-KNN model provided by the present invention.

[0042] Figure 3 It is an intention recognition flow chart of the TGWOA-KNN model provided by the present invention.

[0043] Figure 4 It is a schematic diagram of a dynamic balance training action provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0044] 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 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.

[0045] An embodiment of the present invention discloses a mirror dynamic balance training method for hemiplegic patients. As Figure 1 shown, it includes:

[0046] Collect the muscle contraction change data of the healthy lower limb of the patient;

[0047] Preprocess the collected muscle contraction change data;

[0048] Perform feature data fusion on the preprocessed data, input it into the TGWOA-KNN model for classification, and identify the patient's action intention;

[0049] The affected exoskeleton assists the affected lower limb of the patient to mirror the movement of the healthy lower limb according to the patient's action intention recognized by the TGWOA-KNN model.

[0050] In order to enhance the stability of the patient's core muscles and promote the coordinated recovery of the bilateral lower limb motor muscles, the present invention collects the muscle contraction change signals and posture signals of the healthy side of the subject, and the signals are obtained by FMG sensors and IMU sensors respectively; extracts and fuses the two signal features, sends them into the TGWOA-KNN model for intention recognition, and calculates the healthy joint angle and limb movement height; maps the healthy joint angle to the movement angle of the affected exoskeleton, and maps the limb movement height to the lifting height of the visual display; the subject performs squatting and standing up actions according to the visual prompt and with the assistance of the exoskeleton to complete the visual dynamic balance training.

[0051] Further, the collection of the muscle contraction change data of the healthy lower limb of the patient includes: collecting the muscle contraction change data of the healthy lower limb of the patient through a wearable FMG sensor array; and collecting the posture data of the healthy lower limb of the patient through an IMU sensor.

[0052] Further, the preprocessing of the collected muscle contraction change data includes: processing the collected FMG signals and IMU signals using a Butterworth low-pass filter, then extracting the FMG and IMU feature data, and normalizing the FMG feature data.

[0053] Further, the feature data fusion of the preprocessed data includes: fusing the two types of data using a feature-level fusion method.

[0054] Further, the steps for establishing the TGWOA-KNN model include: optimizing the K value of the KNN model using an improved WOA optimization algorithm; initializing the population using the Tent mapping strategy; and increasing the population diversity using a Gaussian mutation operator.

[0055] Further, the population initialization function generated by using the Tent mapping strategy to initialize the population is:

[0056]

[0057] In the formula, N is the current iteration number, t is the chaotic sequence at the current iteration number, d is the serial number of the chaotic mapping sequence, X max is the upper boundary, X min is the lower boundary;

[0058] The position update function after increasing the population diversity using the Gaussian mutation operator is:

[0059]

[0060] In the formula, X best is the optimal solution of the current algorithm, and G(0,1) is a Gaussian distribution random vector.

[0061] Further, after performing feature data fusion on the preprocessed data, it also includes:

[0062] Using a complementary filter to fuse and calculate the angle:

[0063]

[0064] In the formula, θ filtered is the filtered angle, α is the weight of the filter, θ g is the gyroscope angle, ω is the angular velocity measured by the gyroscope, Δt is the time interval, θ a is the accelerometer angle, θ t-1 is the gyroscope angle at the previous moment, and a1, a2, and a3 are the measured values of the accelerometer on the axes of the space rectangular coordinate system;

[0065] Obtaining the healthy-side joint angle:

[0066]

[0067] In the formula, θ0 is the reference point, θ T is the thigh angle information, θ C is the calf angle information, θ H and θ K are the hip and knee joint angles, which are simultaneously mapped to the motion angles of the affected-side exoskeleton.

[0068] Further, it further includes: calculating the information between multi-point IMU signals to obtain the squatting height:

[0069] The calculation formula for the height change of the IMU sensor in space is:

[0070] h = h t-1 + v cur ·Δt;

[0071] In the formula, h t-1 is the height of the IMU sensor at the previous moment, v cur is the current velocity obtained by integrating the accelerometer, Δt is the time interval, and h is the squatting height of the current limb.

[0072] In a specific embodiment, a mirror dynamic balance training system for hemiplegic patients includes:

[0073] Acquisition module: used to acquire the muscle contraction change data of the healthy lower limb of the patient;

[0074] Preprocessing module: used to preprocess the acquired muscle contraction change data;

[0075] Recognition module: used to perform feature data fusion on the preprocessed data, input it into the TGWOA-KNN model for classification, and recognize the action intention of the patient;

[0076] Execution module: used to control the affected exoskeleton to assist the affected lower limb of the patient to mirror the action of the healthy lower limb according to the action intention of the patient recognized by the TGWOA-KNN model.

[0077] The present invention proposes a TGWOA algorithm to optimize the parameter K of the KNN algorithm to improve the recognition and classification efficiency; establish an intention recognition scheme for the fusion of FMG sensors and IMU sensors, as well as an attitude mapping scheme based on IMUs, and apply them simultaneously to the dynamic balance training of hemiplegic patients; the affected lower limb of the subject completes the same flexion and extension actions as the healthy side with the assistance of the exoskeleton, so as to realize the mirror replication of the movement posture of the healthy lower limb; in addition, visual display is used to guide the actions of the subject to improve the rehabilitation enthusiasm of the subject. Specifically, the schematic diagram of the dynamic balance training action (taking the right side of the subject as an example) is as Figure 4 shown.

[0078] In a specific embodiment, a mirror dynamic balance training method for hemiplegic patients, as Figure 3 shown, the specific steps are as follows:

[0079] S1: Obtain the muscle contraction change data of the healthy lower limb of the subject, which is collected by the wearable FMG sensor array, and obtain the posture data of the healthy lower limb of the subject, which is collected by the IMU sensor;

[0080] S2: Process the collected FMG signals and IMU signals using a Butterworth low-pass filter, then extract the FMG and IMU feature data therefrom, and normalize the FMG feature data;

[0081] S3: Adopt a feature-level fusion method to fuse the two types of data, and send them into the TGWOA-KNN model for classification to identify the patient's motion intention;

[0082] S4: Calculate the information between the multi-point IMU signals, and map the calculated values to the hip and knee joint angles of the affected-side exoskeleton and the lifting height of the mapped character in the visual display respectively;

[0083] S5: After the TGWOA-KNN model confirms that the subject's intention is to squat and stand up, the affected-side exoskeleton assists the affected lower limb of the subject to mirror the movement of the healthy lower limb, and finally completes the squat and stand up movement;

[0084] S6: The subject observes the visual prompt information and repeats the aforementioned actions until the training ends.

[0085] In a specific embodiment, as Figure 2 shown, in the said S3, the classification process of the TGWOA-KNN model is the same as that of the KNN model, and the difference lies in the training process of the KNN model. Since too small a value of K in the KNN model will lead to overfitting, and too large a value of K may introduce too much noise, the K value parameter is particularly important. For this problem, this embodiment uses an improved WOA optimization algorithm to select the optimal K value to solve it. The specific improvement points of the WOA optimization algorithm are as follows:

[0086] (1) First, use the Tent mapping strategy to initialize the population to improve the convergence efficiency of the algorithm. The formula is as follows:

[0087]

[0088] In the formula, the parameter μ ∈ (0, 1), and the population initialization function Xi,d generated by the above mapping strategy is:

[0089]

[0090] In the formula, N is the current iteration number, t is the chaotic sequence at the current iteration number, d is the serial number of the chaotic mapping sequence, X max is the upper boundary, and X min is the lower boundary.

[0091] (2) Second, use the Gaussian mutation operator to increase the population diversity and reduce the possibility of falling into local optimization. The formula is as follows:

[0092]

[0093] Wherein, σ2 is the variance of individuals in the population, σ takes the value of 1, and the value of α is in the range of [0,1]. The position update function improved by the Gaussian mutation operator is as follows:

[0094]

[0095] Wherein, X best is the optimal solution of the current algorithm, and G(0,1) is a Gaussian distribution random vector.

[0096] In a specific embodiment, after the healthy-side attitude data is collected by the IMU sensor in S1, the complementary filter is used to fuse and calculate the angle, and its formula is:

[0097]

[0098] Wherein, θ filtered is the filtered angle, α is the weight of the filter, θ g is the gyroscope angle, ω is the angular velocity measured by the gyroscope, Δt is the time interval, θ a is the accelerometer angle, θ t-1 is the gyroscope angle at the previous moment, and a1, a2, and a3 are the measured values of the accelerometer on the axes of the space rectangular coordinate system. Thus, the healthy-side joint angle is obtained, and its formula is:

[0099]

[0100] Wherein, θ0 is the reference point, θ T is the thigh angle information, θ C is the calf angle information, θ H and θ k are the hip and knee joint angles, and are simultaneously mapped to the movement angles of the affected-side exoskeleton.

[0101] In addition, the calculation formula for the height change of the IMU sensor in space is:

[0102] h = h t-1 + v cur ·Δt;

[0103] Wherein, h t-1 is the height of the IMU sensor at the previous moment, v cur is the current velocity obtained by integrating the accelerometer, Δt is the time interval, and h is the squatting and standing height of the current limb.

[0104] Furthermore, multiplying this height by the mapping coefficient λ gives the lifting height on the visualization display screen.

[0105] In a specific embodiment, the visualization display rule is as follows: control the mapped character in the visualization to avoid obstacles. More specifically, the mapped character in the visualization moves from left to right, and the obstacles move from right to left. The subject needs to control the height of the mapped character in the visualization on the screen by squatting and standing up actions to avoid the obstacles.

[0106] In a specific implementation manner, it further includes: a mirror dynamic balance training device for hemiplegic patients, including: a sensor module, a lower computer module, an upper computer module, and a lower limb rehabilitation exoskeleton module.

[0107] The sensor module includes an FMG sensor and an IMU sensor. The FMG sensor is composed of multiple thin-film pressure sensors and is strapped to the quadriceps femoris, hamstring muscles, etc. of the healthy lower limb of the patient in the form of a circular strap, mainly for detecting the muscle contraction information of the healthy lower limb of the patient. The IMU sensor is mainly used for detecting the attitude information of the healthy hip, thigh, and calf of the patient.

[0108] The lower computer module is mainly used for digitizing the analog signals detected by the sensor module and sending the data to the upper computer module.

[0109] The upper computer module includes a processing program and a visualization display rule. The processing program is mainly used for processing sensor data, including identifying action patterns according to sensor information, calculating the flexion and extension angle data of the healthy joint and the squatting and standing height data of the limb, sending the angle to the lower limb rehabilitation exoskeleton module of the affected side of the subject, and at the same time mapping the height to the lifting height of the mapped character in the visualization display. In addition, the visualization display transmits information to the subject through visual feedback to guide and motivate the subject to complete the dynamic balance training.

[0110] The lower limb rehabilitation exoskeleton module is used to receive the control signal sent by the upper computer, so as to assist the affected lower limb of the subject to mirror the actions of the healthy lower limb and finally complete the squatting and standing actions.

[0111] The present invention combines a lower limb rehabilitation exoskeleton and proposes a rehabilitation training method in which the exoskeleton assists the affected lower limb to mirror the healthy lower limb to perform squatting and standing actions, so as to strengthen the core muscles of the patient and promote the coordinated movement of the bilateral lower limb muscle groups, thereby correcting the wrong compensation mode during the patient's movement. Based on the use of FMG signal recognition, the present invention adds IMU signals to improve the recognition rate and improves the training model by means of algorithm optimization to improve the accuracy of classification and recognition. In addition, the present invention uses an IMU sensor to establish a healthy side attitude mapping scheme to provide control information for the affected side exoskeleton and the mapped character in the visualization display. The present invention guides and motivates the subject to train through the visualization display, improving the enthusiasm of the patient for active rehabilitation.

[0112] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. For the same or similar parts among the various embodiments, reference can be made to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description in the method section.

[0113] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. A mirror dynamic balance training method for hemiplegic patients, characterized in that: include: Collect the muscle contraction change data of the patient's healthy lower limb; Preprocessing the collected muscle contraction change data; The preprocessed data is subjected to feature data fusion and input into the TGWOA-KNN model for classification to identify the patient's action intention; The affected exoskeleton assists the patient's affected lower limb to mirror the healthy lower limb's movements based on the patient's movement intentions identified by the TGWOA-KNN model.

2. The mirror dynamic balance training method for hemiplegic patients according to claim 1, characterized in that: The collecting of the muscle contraction change data of the healthy lower limb of the patient includes: collecting the muscle contraction change data of the healthy lower limb of the patient through a wearable FMG sensor array; and collecting the posture data of the healthy lower limb of the patient through an IMU sensor.

3. The mirror dynamic balance training method for hemiplegic patients according to claim 1, characterized in that: The preprocessing of the collected muscle contraction change data includes: processing the collected FMG signal and IMU signal using a Butterworth low-pass filter, then extracting the FMG and IMU feature data therefrom, and normalizing the FMG feature data.

4. The mirror dynamic balance training method for hemiplegic patients according to claim 1, characterized in that: The fusing of feature data on the preprocessed data includes: fusing two types of data using a feature-level fusion method.

5. The mirror image dynamic balance training method for hemiplegic patients according to claim 1, characterized in that: The steps of establishing the TGWOA-KNN model include: optimizing the K value of the KNN model using an improved WOA optimization algorithm; initializing the population using a Tent mapping strategy; and increasing the diversity of the population using a Gaussian mutation operator.

6. The mirror dynamic balance training method for hemiplegic patients according to claim 5, characterized in that: The population initialization function generated by using the Tent mapping strategy to initialize the population is: Where N is the current iteration number, t is the chaotic sequence under the current iteration number, d is the sequence number of the chaotic mapping sequence, and X max is the upper boundary, X min is the lower boundary; The position update function after using the Gaussian mutation operator to increase population diversity is: Where, X best is the optimal solution of the current algorithm, and G(0,1) is a Gaussian distributed random vector.

7. The mirror image dynamic balance training method for hemiplegic patients according to claim 1, characterized in that: After the feature data fusion is performed on the pre-processed data, the following further comprises: Compute the angle using complementary filter fusion: In the formula, θ filtered is the filtered angle, α is the filter weight, θ g is the gyroscope angle, ω is the angular velocity measured by the gyroscope, Δt is the time interval, θ a is the accelerometer angle, θ t-1 is the gyroscope angle at the previous moment, a1, a2, a3 are the measured values ​​of the accelerometer on the axes of the spatial rectangular coordinate system; Get the healthy side joint angle: In the formula, θ0 is the reference point, θ T is the thigh angle information, θ C is the angle information of the lower leg, θ H and θ K It is the angle of the hip and knee joints, and is also mapped to the movement angle of the exoskeleton on the affected side.

8. The mirror image dynamic balance training method for hemiplegic patients according to claim 7, characterized in that: Also includes: Calculate the information between multi-point IMU signals to get the squat height: The calculation formula for the height change of the IMU sensor in space is: h=h t-1 +v cur ·Δt; In the formula, h t-1 is the height of the IMU sensor at the last moment, v cur is the current speed obtained after integration of the accelerometer, Δt is the time interval, and h is the squat height of the current limb.

9. A mirror dynamic balance training system for hemiplegic patients, characterized in that: include: Acquisition module: used to collect muscle contraction change data of the patient's healthy lower limb; Preprocessing module: used to preprocess the collected muscle contraction change data; Recognition module: used to fuse the feature data of the preprocessed data, input it into the TGWOA-KNN model for classification, and identify the patient's action intention; Execution module: used to control the affected exoskeleton to assist the patient's affected lower limb to mirror the healthy lower limb's movements according to the patient's movement intentions identified by the TGWOA-KNN model.

Images