WRIST REHABILITATION TRAINING SYSTEM BASED ON MUSCLE COORDINATION AND VARIABLE STIFFNESS IMPEDANCE CONTROL

Abstract

A wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control includes the following modules: an electromyographic signal collection and preprocessing module, a muscle co-decomposition and mapping model obtaining module, a man-machine interactive control module, and a virtual reality serious game module; collects a surface electromyographic signal of a forearm of a user, obtains time-domain coordination through non-negative matrix factorization, establishes a position and stiffness estimation model, and controls motion of a target in a serious game through variable stiffness impedance control, so as to complete a training task.

Claims

1. A wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control, comprising: an electromyographic signal collection and preprocessing module configured to collect and preprocess surface electromyographic signals during motion by means of a surface electromyographic signal sensor; a muscle co-decomposition and mapping model obtaining module configured to establish a surface electromyographic signal and joint angle mapping model and a surface electromyographic signal and human stiffness mapping model through a muscle co-decomposition method and a regression network model; a man-machine interactive control module configured to control an agent in a virtual environment through an impedance control method, wherein stiffness in impedance control changes with stiffness of a human arm; and a virtual reality serious game module configured to generate serious game difficulty and virtual interference suitable for rehabilitation training of a user and provide visual feedback.

2. The wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control according to claim 1, wherein the electromyographic signal collection and preprocessing module is configured to: collect the surface electromyographic signals, specifically, use a Delsys wireless surface electromyographic collection device to collect surface electromyographic signals of a flexor carpi radialis muscle, an extensor carpi radialis muscle, a flexor carpi ulnaris muscle, an extensor carpi ulnaris muscle, a flexor digitorum muscle and an extensor digitorum muscle of a human body; and preprocess the surface electromyographic signals, wherein a processing method comprises five steps of full-wave rectification, low-pass filtering, normalization, nerve activation degree extraction, and muscle activation degree extraction.

3. The wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control according to claim 1, wherein the muscle co-decomposition and mapping model obtaining module is configured to: conduct muscle co-decomposition, specifically, conduct a non-negative matrix factorization method on a preprocessed signal, so as to obtain space-domain coordination and time-domain coordination, wherein the non-negative matrix factorization method is as follows:
Z(t)=W.Math.C(t) wherein Z(t) is a preprocessed surface electromyographic signal, W is space-domain coordination, C(t) is space-domain coordination, and all elements in the non-negative matrix factorization method are non-negative; obtain a joint angle mapping model, specifically, train a multi-layer perception model by using time-domain coordinative data and motion tag data, so as to obtain the joint angle mapping model:
x(t)=MLP(W.sup.+.Math.Z(t)) wherein x(t) is a joint angle, W.sup.+ is an inverse matrix of time-domain coordination, Z(t) is the preprocessed surface electromyographic signal, and MLP(.Math.) is a trained multi-layer perception model; and obtain a human stiffness mapping model, specifically, use a coordination effect model to project the preprocessed surface electromyographic signal to coordination effects with different freedom degrees and different directions, wherein the coordination effects are contributions of muscle contraction in all directions, and further obtain, according to a co-contraction principle, a model of the surface electromyographic signal to human stiffness:
K.sub.i(t)=min(W.sub.i+.sup.T.Math.Z(t),W.sub.i−.sup.T.Math.Z(t)) wherein K.sub.i(t) is a stiffness value of each freedom degree, W.sub.i+.sup.T and W.sub.i−.sup.T are coordination effect coefficient matrices in two directions of the freedom degree, and Z(t) is the preprocessed surface electromyographic signal.

4. The wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control according to claim 1, wherein the man-machine interactive control module uses a variable stiffness impedance control method to control the agent in the virtual environment, virtual impedance control involves a damping-spring-mass system, and an expression method of the system is:
M{umlaut over (x)}(t)+B{dot over (x)}(t)+Kx(t)=F.sub.ext wherein M is an object mass, B is a damping coefficient, K is a spring (stiffness) coefficient, x(t), {dot over (x)}(t) and {umlaut over (x)}(t) are a current position, a current speed and a current acceleration, respectively, and F.sub.ext is an extra interference force; in the method, x(t) can be obtained according to a surface electromyographic signal and joint angle relationship model, and {dot over (x)}(t) and {umlaut over (x)}(t) are a first-order derivative and a second-order derivative of x(t), respectively; and K can be obtained according to the model of the surface electromyographic signal to the human stiffness, a relationship between the damping coefficient and the spring (stiffness) coefficient of the system is B=2√{square root over (K)}, the object mass M is a mass of a target preset in a training game, and the extra interference force F.sub.ext is a variable virtual force set in the training game.

5. The wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control according to claim 1, wherein the virtual reality serious game module comprises: a serious game generation module configured to generate different rehabilitation training serious games in different training stages on the basis of a pygame platform; a feedback module configured to provide a patient with game information such as interference in the virtual environment and a motion position of a virtual agent; and a serious game adjustment module configured to adjust the serious game difficulty according to an effect of each cycle of rehabilitation training, so as to increase challenges of rehabilitation training.

6. A method for using the wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control according to claim 1, comprising: firstly, collecting and preprocessing surface electromyographic signals during flexion and extension of wrists and fingers of dominant limbs of a patient, then determining a surface electromyographic signal and joint angle relationship model and a surface electromyographic signal and human stiffness relationship model through a muscle coordination method, then setting serious game scene and difficulty, then collecting and preprocessing surface electromyographic signals of non-dominant limbs of the patient, then obtaining a joint angle and human stiffness according to the surface electromyographic signal and joint angle relationship model and the surface electromyographic signal and human stiffness relationship model, and finally, controlling an agent in a virtual environment to complete a rehabilitation training task through a variable stiffness impedance control method.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0034] FIGURE is a block diagram of a system of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] The present invention will be further described with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0036] A rehabilitation training task in the present invention generally refers to free motion of wrists and fingers in daily lives, such as taking things, doing housework or other complex tasks. Herein, a therapist/technician informs a subject of a game target, and the subject may control a target to move through actions of fingers and wrists, so as to complete the task.

[0037] As shown in FIGURE, a wrist rehabilitation training system based on muscle coordination and variable stiffness impedance control according to an example of the present invention is used through the following steps:

[0038] (1) Surface electromyographic signals during flexion and extension of wrists and fingers are collected and preprocessed.

[0039] The surface electromyographic signals are interactive information most commonly used in the rehabilitation training system. The surface electromyographic signals contain a large amount of action and stiffness information, and the information may be decomposed through specific algorithms and used to control an artificial limb, a robot, or a target in a virtual scene. A Delsys wireless surface electromyographic collection system is used to collect bioelectric signals generated during muscle motion, a subject needs to conduct flexion and extension of wrists and fingers separately, and surface electromyographic signals of a flexor carpi radialis muscle, an extensor carpi radialis muscle, a flexor carpi ulnaris muscle, an extensor carpi ulnaris muscle, a flexor digitorum muscle and an extensor digitorum muscle are recorded. Operations such as full-wave rectification and low-pass filtering are conducted on the collected surface electromyographic signals, so as to obtain an envelope curve of the electromyographic signals, then normalization is conducted such that processed data is within a range of 0-1, then a nerve activation degree is obtained through a nerve activation model, and finally a muscle activation degree is obtained through a muscle activation model.

[0040] (2) A surface electromyographic signal and joint angle relationship model is determined.

[0041] According to a muscle co-decomposition principle, the preprocessed electromyographic signals are decomposed through a non-negative matrix factorization method. Muscle co-decomposition is conducted on signals generated during flexion and extension of wrists and fingers separately.

Z.sub.wf(t)=W.sub.wf.Math.C.sub.wf(t)

Z.sub.we(t)=W.sub.we.Math.C.sub.we(t)

Z.sub.ff(t)=W.sub.ff.Math.C.sub.ff(t)

Z.sub.fe(t)=W.sub.fe.Math.C.sub.fe(t)

[0042] Z.sub.wf(t), Z.sub.we(t), Z.sub.ff(t) and Z.sub.fe(t) are electromyographic signals during wrist flexion, wrist extension, finger flexion and finger extension, respectively, W.sub.wf, W.sub.we, W.sub.ff and W.sub.fe are space-domain coordination effects corresponding to wrist flexion, wrist extension, finger flexion and finger extension, respectively, and C.sub.wf(t), C.sub.we(t), C.sub.ff(t) and C.sub.fe(t) are time-domain coordination effects corresponding to wrist flexion, wrist extension, finger flexion and finger extension, respectively. The obtained space-domain coordination effects are combined in columns, so as to obtain overall space-domain coordination:

W=[W.sub.wf,W.sub.we,W.sub.ff,W.sub.fe]

[0043] A multi-layer perception model regression network is trained by using obtained time-domain coordination and tag data, so as to obtain a relationship module of an electromyographic signal to a joint angle:

x(t)=MLP(W.sup.+.Math.Z(t))

[0044] where x(t) is a joint angle, W.sup.+ is an inverse matrix of time-domain coordination, Z(t) is the preprocessed surface electromyographic signal, and MLP(.Math.) is a trained multi-layer perception model.

[0045] (3) A surface electromyographic signal and human stiffness model is determined.

[0046] The preprocessed surface electromyographic signal is projected to coordination effects with different freedom degrees and different directions according to a coordination effect model, where the coordination effects are contributions of muscle contraction in all directions, and a specific projection method is the same as the muscle co-decomposition method in (2). According to a co-contraction principle, a model of a surface electromyographic signal to human stiffness is obtained as follows:

K.sub.w(t)=min(W.sub.wf.sup.T.Math.Z(t),W.sub.we.sup.T.Math.Z(t))

K.sub.f(t)=min(W.sub.ff.sup.T.Math.Z(t),W.sub.fe.sup.T.Math.Z(t))

[0047] where K.sub.w(t) and K.sub.f(t) are stiffness in a wrist flexion and extension direction and stiffness in a finger flexion and extension direction, respectively W.sub.wf.sup.T, W.sub.we.sup.T, W.sub.ff.sup.T and W.sub.fe.sup.T are coordination effect coefficient matrices in wrist flexion, wrist extension and finger flexion directions, respectively, and Z(t) is the preprocessed surface electromyographic signal.

[0048] (4) A task control scene is initialized and task complexity is set.

[0049] In order to improve enthusiasm of a subject in participating in rehabilitation training, two games, flying bird and captain rogers, are used in different training game scenes, and the game complexity is changed by increasing or decreasing obstacles in the game.

[0050] (5) Rehabilitation training is conducted.

[0051] Through multi-time long-term training, the subject may undergo effective rehabilitation.

[0052] A virtual impedance control method is used in control:

M{umlaut over (x)}(t)+B{dot over (X)}(t)+Kx(t)=F.sub.ext

[0053] where M is an object mass, B is a damping coefficient, K is a spring (stiffness) coefficient, x(t), {dot over (x)}(t) and {umlaut over (x)}(t) are a current position, a current speed and a current acceleration, respectively, and F.sub.ext is an extra interference force. A joint angle obtained according to the surface electromyographic signal and joint angle relationship model is mapped to coordinates of a control target of a training game through projection transformation. Human stiffness obtained according to the surface electromyographic signal and human stiffness model is taken as the spring (stiffness) coefficient of the virtual impedance control method. The subject needs to control actual output targets so as to complete related tasks.

[0054] (6) The task scene and task complexity are adjusted according to game performance.

[0055] During rehabilitation training, time and a score of each task completed by the subject may be recorded. The task scene and task complexity are adjusted according to the performance after each cycle of training. If the time to complete the game decreases or the score increases, the game difficulty is increased, and otherwise, the game difficulty is decreased.

[0056] It should be noted that the above description only illustrates the technical idea of the present invention, and cannot be used to limit the protection scope of the present invention. Those of ordinary skill in the art can also make some improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also fall within the protection scope of claims of the present invention.