CN105031812A

CN105031812A - Functional electrostimulation closed-loop control system and method of electromyographic signal feedback

Info

Publication number: CN105031812A
Application number: CN201510313665.2A
Authority: CN
Inventors: 李展; 程洪; 殷紫光
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2015-06-09
Filing date: 2015-06-09
Publication date: 2015-11-11

Abstract

The invention discloses a functional electric stimulation closed-loop control system and method for myoelectric signal feedback. First, after the functional electric stimulator completes initialization and signal synchronization, it begins to electrically stimulate the controlled object; the myoelectric signal collector collects The original myoelectric signal generated when the muscle is electrically stimulated is preprocessed; then the Hammerstein model with time delay is used to establish the muscle contraction model, and the Kalman filter method is used for parameter identification; finally, the online real-time prediction of the muscle contraction model is performed. , calculate the optimal electric pulse width control amount, and feed it back to the functional electric stimulator to update its parameters to realize real-time adaptive control. The invention completes the closed-loop control and adjustment of the pulse number of electrical stimulation by calculating the absolute average amplitude of the electromyographic signal of the controlled object; improves the control accuracy of the closed-loop functional electrical stimulation system, and realizes the functionality of electromyographic signal feedback Adaptive control of electrical stimulation.

Description

A functional electrical stimulation closed-loop control system and method for electromyographic signal feedback

技术领域 technical field

本发明涉及功能性电刺激领域，特别是涉及一种肌电信号反馈的功能性电刺激闭环控制系统及方法。 The invention relates to the field of functional electrical stimulation, in particular to a functional electrical stimulation closed-loop control system and method for feedback of myoelectric signals .

背景技术 Background technique

目前的商用和临床使用的功能性电刺激控制系统绝大部分采用开环控制系统方法设计，并且没有引入肌电信号作为反馈对被控体进行模型和控制器更新。目前流行的开环原理的控制系统往往缺少自适应调节能力，不能精确完成目标肌肉缩张控制任务而无法实现相应的康复控制方案，其具体缺陷如下。 Most of the current commercial and clinical functional electrical stimulation control systems are designed using the open-loop control system method, and the EMG signal is not introduced as feedback to update the model and controller of the controlled body. The current popular open-loop control system often lacks the ability of self-adaptive adjustment, and cannot accurately complete the target muscle contraction control task and cannot realize the corresponding rehabilitation control plan. The specific defects are as follows.

1、缺少肌电信号的反馈信息对控制器和被控对象建模的自适应调节，控制器和模型辨识误差较大； 1. Lack of feedback information of electromyographic signals to adaptively adjust the modeling of the controller and the controlled object, and the identification error of the controller and the model is relatively large;

2、闭环控制器选择范围太窄，控制器需过多依据经验调节； 2. The selection range of the closed-loop controller is too narrow, and the controller needs to be adjusted too much based on experience;

3、不能通过肌电信号反馈信息更新或更改电刺激电极的安置位置。 3. The placement position of the electrical stimulation electrodes cannot be updated or changed through the feedback information of the electromyography signal.

发明内容 Contents of the invention

本发明的目的在于克服现有技术的不足，提供一种肌电信号反馈的功能性电刺激闭环控制系统及方法，通过计算被刺激肌肉的肌电信号的绝对平均幅值，完成对电刺激的脉冲数的闭环控制调节；建立了电刺激脉冲数-肌电绝对平均幅值模型，该肌肉模型基于含时滞项Hammerstein结构；改进了目前闭环功能性电刺激系统控制精度和解决了其控制系统参数自适应能力不足的问题，实现了肌电信号反馈的功能性电刺激自适应控制。 The purpose of the present invention is to overcome the deficiencies of the prior art, to provide a functional electrical stimulation closed-loop control system and method for myoelectric signal feedback , and to complete the electrical stimulation by calculating the absolute average amplitude of the myoelectric signal of the stimulated muscle. The closed-loop control adjustment of the number of pulses; established the electrical stimulation pulse number-absolute average amplitude model of myoelectricity, the muscle model is based on the Hammerstein structure with time-delay items; improved the control accuracy of the current closed-loop functional electrical stimulation system and solved its control system Insufficient parameter self-adaptive ability, realized functional electrical stimulation self-adaptive control of EMG signal feedback.

本发明的目的是通过以下技术方案来实现的：一种肌电信号反馈的功能性电刺激闭环控制系统，所述系统包括以下多个模块： The purpose of the present invention is achieved by the following technical solutions: a functional electrical stimulation closed-loop control system for myoelectric signal feedback, said system comprising the following modules:

功能性电刺激器，采用脉冲电流刺激被控对象的肌肉。 A functional electrical stimulator uses pulsed currents to stimulate the muscles of the subject.

肌电信号采集器，捕捉电刺激肌肉时产生的原始肌电信号，并对原始肌电信号进行预处理，提取出电刺激脉冲数、电脉冲宽度和肌电绝对平均幅值。 The myoelectric signal collector captures the original myoelectric signal generated when the muscles are electrically stimulated, and preprocesses the original myoelectric signal to extract the number of electrical stimulation pulses, the width of the electrical pulse and the absolute average amplitude of the myoelectricity.

肌肉缩张模型，采用含时滞项的Hammerstein模型对电刺激脉冲数和肌电绝对平均幅值进行建模，采用卡尔曼滤波的方式对肌肉模型进行参数辨识，得到模型的线性和非线性两部分的系数。 For the muscle contraction model, the Hammerstein model with time-delay items is used to model the number of electrical stimulation pulses and the absolute average amplitude of myoelectricity, and the Kalman filter is used to identify the parameters of the muscle model. part of the coefficient.

模型预测控制器，根据参考肌肉激励轨迹和模式及参数辨识结果，通过最优预测控制算法确定脉冲电流的最优的电脉冲宽度的控制量，再将该控制量反馈到功能性电刺激器，对功能性电刺激器的进行参数更新，使得功能性电刺激器输出所需的脉冲电流，并对功能性电刺激器进行实时的自适应控制。 The model predictive controller, according to the reference muscle excitation trajectory and the pattern and parameter identification results, determines the control amount of the optimal electric pulse width of the pulse current through the optimal predictive control algorithm, and then feeds back the control amount to the functional electric stimulator, The parameters of the functional electrical stimulator are updated, so that the functional electrical stimulator can output the required pulse current, and the functional electrical stimulator can be adaptively controlled in real time.

所述原始肌电信号为复合肌肉动作电位CMAP，即M波。 The original myoelectric signal is a compound muscle action potential CMAP, that is, an M wave.

一种肌电信号反馈的功能性电刺激闭环控制方法，所述方法包括一下多个步骤： A functional electrical stimulation closed-loop control method for electromyographic signal feedback, said method comprising the following multiple steps:

S1：功能性电刺激器初始化及信号同步； S1: Functional electrical stimulator initialization and signal synchronization;

S2：肌电信号采集器采集电刺激肌肉时产生的原始肌电信号，并对原始肌电信号进行预处理，提取出电刺激脉冲数、电脉冲宽度和肌电绝对平均幅值； S2: The myoelectric signal collector collects the original myoelectric signal generated when the muscle is electrically stimulated, and preprocesses the original myoelectric signal to extract the number of electrical stimulation pulses, the width of the electrical pulse and the absolute average amplitude of the myoelectricity;

S3：采用含时滞项的Hammerstein模型对电刺激脉冲数和肌电绝对平均幅值进行建模，采用卡尔曼滤波的方式对肌肉模型进行参数辨识； S3: The number of electrical stimulation pulses and the absolute average amplitude of myoelectricity are modeled using the Hammerstein model with a time-delay term, and the parameters of the muscle model are identified by means of Kalman filtering;

S4：根据参考肌肉激励轨迹和模式及参数辨识结果，通过最优预测控制算法确定脉冲电流的最优的电脉冲宽度的控制量，再将该控制量反馈到功能性电刺激器，对功能性电刺激器的进行参数更新，使得功能性电刺激器输出所需的脉冲电流，并对功能性电刺激器进行实时的自适应控制。 S4: According to the reference muscle excitation trajectory and the pattern and parameter identification results, determine the control amount of the optimal electric pulse width of the pulse current through the optimal predictive control algorithm, and then feed back the control amount to the functional electrical stimulator. The parameters of the electrical stimulator are updated, so that the functional electrical stimulator outputs the required pulse current, and real-time adaptive control is performed on the functional electrical stimulator.

步骤S2中所述的预处理包括： The preprocessing described in step S2 includes:

①去畸变量处理：在一个电刺激刺激周期内设定阈值模版做幅值判别； ① De-distortion processing: set the threshold template for amplitude discrimination within one electrical stimulation cycle;

②提取原始肌电信号的绝对值和平均值；得到可易于控制和辨识的肌电信号幅值； ② Extract the absolute value and average value of the original EMG signal; obtain the EMG signal amplitude that can be easily controlled and identified;

③计算平均值需要窗宽度：窗宽度为肌电信号的采样频率和电刺激脉冲电流频率之比的四舍五入近似整数值； ③ Calculating the average value requires a window width: the window width is the rounded approximate integer value of the ratio between the sampling frequency of the EMG signal and the frequency of the electrical stimulation pulse current;

④归一化处理：保证卡尔曼滤波的稳定性。 ④Normalization processing: to ensure the stability of Kalman filtering.

步骤S3中所述参数辨识的计算公式为： The calculation formula for the parameter identification described in step S3 is:

$y the y ((k k)) = = {Σ Σ}_{i i = = 11}^{l l} {a a}_{i i} ((k k)) y the y ((k k - - i i)) + + {Σ Σ}_{i i = = 11}^{m m} {Σ Σ}_{j j = = 00}^{n no} {b b}_{i i} ((k k)) {c c}_{j j} ((k k)) {u u}^{j j} ((k k - - i i))$

式中，y(k)—原始肌电信号的绝对平均值；y(k-i)—后向时肌电信号的绝对平均值； In the formula, y(k)—the absolute average value of the original EMG signal; y(k-i)—the absolute average value of the EMG signal in the backward direction;

u(k)—原始肌电信号的电脉冲宽度；u^j(k-i)—后向电脉冲宽度的j次幂； u(k)—the electrical pulse width of the original EMG signal; u ^j (ki)—the jth power of the backward electrical pulse width;

a_i(k)—待辨识的被控激励系统线性项参数；b_i(k)—待辨识的被控激励系统非线性项参数；c_j(k)—待辨识的被控激励系统非线性项参数； a _i (k)—parameters of the linear term of the controlled excitation system to be identified; b _i (k)—parameters of the nonlinear term of the controlled excitation system to be identified; c _j (k)—parameters of the nonlinear term of the controlled excitation system to be identified item parameter;

k—电刺激迭代循环数；i—线性项动态阶数；l—线性项阶数上限；j—幂次项动态阶数；m—非线性项阶数上限；n—非线性项阶数上限； k—electric stimulation iteration cycle number; i—dynamic order of linear item; l—upper limit of linear item order; j—dynamic order of power item; m—upper limit of nonlinear item order; n—upper limit of nonlinear item order ;

—模型线性自回归部分；—模型非线性自回归部分。 — the linear autoregressive part of the model; — The nonlinear autoregressive part of the model.

将步骤S3中的肌肉模型转换为状态空间结构，所述状态空间结构的形式为： The muscle model in step S3 is converted into a state space structure, and the form of the state space structure is:

x(k)＝A(k)x(k-1)+B(k)Φ(u(k-1)) x(k)=A(k)x(k-1)+B(k)Φ(u(k-1))

y(k)＝x₁(k) y(k)=x ₁ (k)

式中，y(k)—原始肌电信号的绝对平均值； In the formula, y(k)—the absolute average value of the original EMG signal;

x(k)—状态方程变量；x(k-1)—上一状态方程变量；x₁(k)—上一状态方程变量首元素； x(k)—state equation variable; x(k-1)—previous state equation variable; x ₁ (k)—previous state equation variable first element;

A(k)—系数矩阵；B(k)—系数矩阵； A(k)—coefficient matrix; B(k)—coefficient matrix;

Φ(u(k-1))—关于电刺激脉冲数的非线性矩阵；u(k-1)—上一电刺激脉冲数； Φ(u(k-1))—Nonlinear matrix about the number of electrical stimulation pulses; u(k-1)—Number of last electrical stimulation pulses;

k—电刺激迭代循环数。 k—number of electrical stimulation iteration cycles.

所述参数辨识包括先验证辨识过程，采用基于遗忘因子的卡尔曼滤波方法进行参数辨识，其先验证辨识过程的公式为： The parameter identification includes first verifying the identification process, using the forgetting factor-based Kalman filter method for parameter identification, and the formula for first verifying the identification process is:

$\overset{^^}{x x} ((k k)) = = F f ((x x ((k k - - 11)),, u u ((k k - - 11))))$

$\overset{^^}{P P} ((k k)) = = A A ((k k - - 11)) P P ((k k - - 11)) {A A}^{T T} ((k k - - 11)) / / λ λ$

$\overset{^^}{y the y} ((k k)) = = {\overset{^^}{x x}}_{11} ((k k))$

式中，—先验辨识状态矩阵；—先验协方差矩阵；—预测控制输出； In the formula, — a priori identification state matrix; — prior covariance matrix; — predictive control output;

x(k-1)—上一状态变量测量值；u(k-1)—上一电刺激脉冲数；F(x(k-1),u(k-1))—上一状态变量测量值和电刺激脉冲数的非线性映射； x(k-1)—the measured value of the last state variable; u(k-1)—the number of the last electrical stimulation pulse; F(x(k-1),u(k-1))—the measured value of the last state variable Non-linear mapping of value and number of electrical stimulation pulses;

A(k-1)—待辨识系数矩阵；A^T(k-1)—系数矩阵转置；u(k-1)—上一电刺激脉冲数； A(k-1)—coefficient matrix to be identified; A ^T (k-1)—transposition of coefficient matrix; u(k-1)—number of last electrical stimulation pulses;

—先验辨识状态矩阵首元素；k—电刺激迭代循环数；T—转置操作；λ—遗忘因子。 —Prior identification of the first element of the state matrix; k—number of electrical stimulation iteration cycles; T—transposition operation; λ—forgetting factor.

步骤S4中所述的参数更新包括后验证更新过程，其后验证更新过程的公式为： The parameter update described in the step S4 includes post-authentication update process, and the formula of the post-authentication update process is:

$S S ((k k)) = = H h ((k k)) \overset{^^}{P P} ((k k)) {H h}^{T T} ((k k)) + + I I$

$K K ((k k)) = = \overset{^^}{P P} ((k k)) {H h}^{T T} ((k k)) {S S}^{- - 11} ((k k))$

$x x ((k k)) &LeftArrow; &LeftArrow; \overset{^^}{x x} ((k k)) + + K K ((k k)) ((y the y ((k k)) - - \overset{^^}{y the y} ((k k))))$

S(k)—卡尔曼滤波系数矩阵；H(k)—卡尔曼滤波系数矩阵；H^T(k)—卡尔曼滤波系数矩阵的转置； S(k)—Kalman filter coefficient matrix; H(k)—Kalman filter coefficient matrix; H ^T (k)—transpose of Kalman filter coefficient matrix;

K(k)—状态更新迭代系数矩阵；S^-1(k)—卡尔曼滤波系数矩阵逆； K(k)—state update iteration coefficient matrix; S ^-1 (k)—inverse Kalman filter coefficient matrix;

x(k)—状态变量；y(k)—原始肌电信号的绝对平均值； x(k)—state variable; y(k)—absolute average value of the original EMG signal;

P(k)—测量值系数矩阵；—预估测量值系数矩阵； P(k)—measured value coefficient matrix; — estimated measured value coefficient matrix;

k—电刺激迭代循环数；T—转置操作；I—单位矩阵；λ—遗忘因子。 k—number of electrical stimulation iteration cycles; T—transpose operation; I—identity matrix; λ—forgetting factor.

根据参考肌肉激励轨迹和模式，对于肌肉模型进行预测控制，选取如下的成本函数，并通过该成本函数对肌肉模型进行预测控制： According to the reference muscle excitation trajectory and mode, the muscle model is predictively controlled, and the following cost function is selected, and the muscle model is predictively controlled through the cost function:

$J J ((k k)) = = {Σ Σ}_{j j = = 11}^{Np Np} {ϵ ϵ}_{j j} [[\overset{^^}{u u} ((k k + + j j | | k k)) - - {u u}_{d d} ((k k + + j j)) {]]}^{22} + + {Σ Σ}_{j j = = 11}^{Nu Nu} {δ δ}_{j j} [[h h ((k k + + j j | | k k)) - - h h ((k k)) {]]}^{22}$

式中，J(k)—成本函数；j|k—电刺激迭代循环的前向预测或控制动态阶数；—前向预测肌电激励幅度； In the formula, J(k)—cost function; j|k—forward prediction or control dynamic order of electrical stimulation iterative cycle; - Forward prediction of myoelectric excitation amplitude;

u_d(k+j)—期望肌肉激励控制目标；ε_j—预测区间的优化系数； u _d (k+j)—desired muscle stimulation control target; ε _j —optimization coefficient of prediction interval;

h(k+j|k—电刺激前向脉冲数的多项式组合；h(k)—电刺激脉冲宽度的控制量调节幅度；δ_j—控制区间的优化系数； h(k+j|k—the polynomial combination of the number of forward pulses of electrical stimulation; h(k)—the adjustment range of the control amount of electrical stimulation pulse width; δ _j —the optimization coefficient of the control interval;

Np—预设预测区间；Nu—预设控制区间； Np—preset prediction interval; Nu—preset control interval;

k—电刺激迭代循环数；j—阶数动态值；d—期望值下角标。 k—the iterative cycle number of electrical stimulation; j—the dynamic value of the order; d—the subscript of the expected value.

本发明的有益效果是： The beneficial effects of the present invention are:

1)在功能性电刺激器完成初始化及信号同步后，开始对被控对象进行电刺激；肌电信号采集器采集电刺激肌肉时产生的原始肌电信号，并进行预处理；再采用含时滞项的Hammerstein模型建立肌肉缩张模型，采用卡尔曼滤波方式进行参数辨识；最后对肌肉缩张模型进行在线实时预测，计算最优的电脉冲宽度的控制量，并反馈到功能性电刺激器，对其进行参数更新，实现实时的自适应控制。 1) After the initialization and signal synchronization of the functional electrical stimulator are completed, the controlled object begins to be electrically stimulated; the myoelectric signal collector collects the original myoelectric signal generated when the muscle is electrically stimulated, and preprocesses it; The Hammerstein model of the hysteresis term establishes the muscle contraction model, and uses the Kalman filter method for parameter identification; finally, the online real-time prediction of the muscle contraction model is performed, and the optimal control amount of the electrical pulse width is calculated, and fed back to the functional electrical stimulator , to update its parameters to realize real-time adaptive control.

2)本发明通过计算被控对象的肌电信号的绝对平均幅值，完成对电刺激的脉冲数的闭环控制调节；建立了电刺激脉冲数-肌电绝对平均幅值的肌肉缩张模型，该肌肉缩张模型基于含时滞项的Hammerstein结构；提高了闭环功能性电刺激系统控制精度，实现了肌电信号反馈的功能性电刺激自适应控制。 2) the present invention completes the closed-loop control adjustment to the pulse number of electrical stimulation by calculating the absolute average amplitude of the myoelectric signal of the controlled object; the muscle contraction and contraction model of the electrical stimulation pulse number-myoelectric absolute average amplitude is established, The muscle contraction and contraction model is based on the Hammerstein structure with time-delay items; it improves the control precision of the closed-loop functional electrical stimulation system, and realizes the adaptive control of functional electrical stimulation with EMG signal feedback.

附图说明 Description of drawings

图1为本发明功能性电刺激闭环控制系统的系统结构框图； Fig. 1 is the system structural block diagram of functional electrical stimulation closed-loop control system of the present invention;

图2为本发明功能性电刺激闭环控制方法的流程图； Fig. 2 is a flow chart of the functional electrical stimulation closed-loop control method of the present invention;

图3为本发明中肌电信号处理的流程示意图； Fig. 3 is the schematic flow sheet of myoelectric signal processing in the present invention;

图4为本发明中参数辨识的流程示意图； Fig. 4 is a schematic flow chart of parameter identification in the present invention;

图5为本发明中功能性电刺激闭环控制流程示意图。 Fig. 5 is a schematic diagram of the closed-loop control flow of functional electrical stimulation in the present invention.

具体实施方式 Detailed ways

下面结合附图进一步详细描述本发明的技术方案，但本发明的保护范围不局限于以下所述。 The technical solution of the present invention will be further described in detail below in conjunction with the accompanying drawings , but the protection scope of the present invention is not limited to the following description.

(一)功能性电刺激闭环控制系统 (1) Functional electrical stimulation closed-loop control system

如图1所示，一种肌电信号反馈的功能性电刺激闭环控制系统，所述系统包括以下多个模块： As shown in Figure 1, a functional electrical stimulation closed-loop control system for myoelectric signal feedback, the system includes the following multiple modules:

①功能性电刺激器，采用脉冲电流刺激被控对象的肌肉。功能性电刺激器输出的电刺激脉冲形状采用双极型。 ① Functional electrical stimulator, using pulse current to stimulate the muscles of the controlled object. The shape of the electrical stimulation pulse output by the functional electrical stimulator is bipolar.

②肌电信号采集器，捕捉电刺激肌肉时产生的原始肌电信号，并对原始肌电信号进行预处理，提取出电刺激脉冲数、电脉冲宽度和肌电绝对平均幅值。 ②Myoelectric signal collector, which captures the original myoelectric signal generated when the muscles are electrically stimulated, and preprocesses the original myoelectric signal to extract the number of electrical stimulation pulses, the width of the electrical pulse and the absolute average amplitude of the myoelectricity.

肌电信号采集器的采样频率根据实际情况设置，来保证其能够完全捕捉电刺激肌肉时产生的M波。 The sampling frequency of the myoelectric signal collector is set according to the actual situation to ensure that it can fully capture the M wave generated when the muscle is electrically stimulated.

③肌肉缩张模型，采用含时滞项的Hammerstein模型对电刺激脉冲数和肌电绝对平均幅值进行建模，采用卡尔曼滤波的方式对肌肉模型进行参数辨识，得到模型的线性和非线性两部分的系数。 ③ Muscle contraction and contraction model, using the Hammerstein model with time-delay items to model the number of electrical stimulation pulses and the absolute average amplitude of myoelectricity, and using the Kalman filter to identify the parameters of the muscle model to obtain the linearity and nonlinearity of the model Coefficient of two parts.

④模型预测控制器，根据参考肌肉激励轨迹和模式及参数辨识结果，通过最优预测控制算法确定脉冲电流的最优的电脉冲宽度的控制量，再将该控制量反馈到功能性电刺激器，对功能性电刺激器的进行参数更新，使得功能性电刺激器输出所需的脉冲电流，并对功能性电刺激器进行实时的自适应控制。 ④The model predictive controller, according to the reference muscle excitation trajectory and the pattern and parameter identification results, determines the control amount of the optimal electric pulse width of the pulse current through the optimal predictive control algorithm, and then feeds back the control amount to the functional electric stimulator , updating the parameters of the functional electrical stimulator, so that the functional electrical stimulator outputs the required pulse current, and performs real-time adaptive control on the functional electrical stimulator.

模型预测控制器采用预测控制模式，包括预测区间和控制区间，还可通过限定预测区间和控制区间的时间段，来保证模型预测控制器具有足够的稳定性。 The model predictive controller adopts the predictive control mode, including the prediction interval and the control interval, and the time period of the prediction interval and the control interval can also be limited to ensure that the model predictive controller has sufficient stability.

(二)功能性电刺激闭环控制方法 (2) Functional electrical stimulation closed-loop control method

如图2所示，一种肌电信号反馈的功能性电刺激闭环控制方法，所述方法包括一下多个步骤： As shown in Figure 2, a functional electrical stimulation closed-loop control method for myoelectric signal feedback, the method includes the following steps:

S1：功能性电刺激器初始化及信号同步。 S1: Functional electrical stimulator initialization and signal synchronization.

安置功能性电刺激器的正负极电极于对应肌肉附近神经组织处，设置功能性电刺激器的脉冲模式为常量幅值模式，观察和检查对应肌电信号的变化情况，如出现M波表明系统信号同步良好。 Place the positive and negative electrodes of the functional electrical stimulator on the nerve tissue near the corresponding muscles, set the pulse mode of the functional electrical stimulator to constant amplitude mode, observe and check the changes of the corresponding myoelectric signals, if M waves appear The system signals are well synchronized.

S2：肌电信号采集器采集电刺激肌肉时产生的原始肌电信号，并对原始肌电信号进行预处理，提取出电刺激脉冲数、电脉冲宽度和肌电绝对平均幅值。 S2: The myoelectric signal collector collects the original myoelectric signal generated when the muscle is electrically stimulated, and preprocesses the original myoelectric signal to extract the number of electrical stimulation pulses, the width of the electrical pulse and the absolute average amplitude of the myoelectricity.

功能性电刺激器的正负极电极根据欧洲肌肉电信号安置标准安置在肌肉表面，接地端一般安置在肌肉附近的露骨处。 The positive and negative electrodes of the functional electrical stimulator are placed on the muscle surface according to the European muscle electrical signal placement standard, and the grounding terminal is generally placed on the exposed place near the muscle.

S3：如图4所示，在完成步骤S1和S2后，设置功能性电刺激器的脉冲模式为随机幅值的梯形模式，采用含时滞项的Hammerstein模型对电刺激脉冲数和肌电绝对平均幅值进行建模，采用卡尔曼滤波的方式对肌肉模型进行参数辨识； S3: As shown in Figure 4, after completing steps S1 and S2, set the pulse pattern of the functional electrical stimulator to a trapezoidal pattern with random amplitudes, and use the Hammerstein model with a time-delay item to analyze the number of electrical stimulation pulses and the absolute value of myoelectricity. The average amplitude is used for modeling, and the parameter identification of the muscle model is carried out by means of Kalman filter;

S4：在完成步骤S3后，设置肌肉缩张模型为参考跟踪模式，根据参考肌肉激励轨迹和模式及参数辨识结果，通过最优预测控制算法确定脉冲电流的最优的电脉冲宽度的控制量，再将该控制量反馈到功能性电刺激器，对功能性电刺激器的进行参数更新，使得功能性电刺激器输出所需的脉冲电流，并对功能性电刺激器进行实时的自适应控制。控制全程保持刺激强度恒定，只调节脉冲宽度。 S4: After completing step S3, set the muscle contraction and tension model as the reference tracking mode, and determine the control amount of the optimal electric pulse width of the pulse current through the optimal predictive control algorithm according to the reference muscle excitation trajectory, mode and parameter identification results, Then feed back the control amount to the functional electrical stimulator, update the parameters of the functional electrical stimulator, make the functional electrical stimulator output the required pulse current, and perform real-time adaptive control on the functional electrical stimulator . The control keeps the stimulation intensity constant throughout the whole process, and only adjusts the pulse width.

如图3所示，步骤S2中所述的预处理包括： As shown in Figure 3, the preprocessing described in step S2 includes:

①去畸变量处理：在一个电刺激刺激周期内设定阈值模版做幅值判别，如肌电信号超过模板阈值，进行归零化作用于得到原始M波； ① De-distortion processing: set the threshold template for amplitude discrimination within one electrical stimulation cycle, and if the EMG signal exceeds the template threshold, perform zeroing to obtain the original M wave;

从该参数辨识公式可以看出，被辨识模型为递归结构，包含了线性和非线性项，便于实时在线实现，所需辨识参数任务明确，将易于实际的功能性电刺激控制器的内核预测控制或PID算法实现。 From the parameter identification formula, it can be seen that the identified model is a recursive structure, including linear and nonlinear items, which is convenient for real-time online implementation, and the required identification parameters are clear, which will be easy to implement in the kernel predictive control of the actual functional electrical stimulation controller. Or PID algorithm implementation.

为了使采用卡尔曼滤波进行系统辨识的在线递归更易于实现，本发明将步骤S3中的肌肉模型转换为状态空间结构，所述状态空间结构的形式为： In order to make the online recursion of system identification using Kalman filter easier to realize, the present invention converts the muscle model in step S3 into a state space structure, and the form of the state space structure is:

x(k)＝A(k)x(k-1)+B(k)Φ(u(k-1)) x(k)=A(k)x(k-1)+B(k)Φ(u(k-1))

y(k)＝x₁(k) y(k)=x ₁ (k)

状态空间结构中，Φ(u(k-1))＝[u(k-1)u²(k-1)…uⁿ(k-1)]，式中，u²(k-1)—电脉冲数的2阶非线性项；uⁿ(k-1)—电脉冲数的n阶非线性项；n—非线性阶数。 In the state space structure, Φ(u(k-1))=[u(k-1)u ² (k-1)…u ⁿ (k-1)], where u ² (k-1)— The second-order nonlinear item of the electric pulse number; u ⁿ (k-1)—the n-order nonlinear item of the electric pulse number; n—the nonlinear order.

$\overset{^^}{x x} ((k k)) = = F f ((x x ((k k - - 11)),, u u ((k k - - 11))))$

$\overset{^^}{y the y} ((k k)) = = {\overset{^^}{x x}}_{11} ((k k))$

x(k-1)—上一状态变量测量值；u(k-1)—上一电刺激脉冲数；F(x(k-1),u(k-1))—上一状态变量测量值和电刺激脉冲数的非线性映射； x(k-1)—the measured value of the last state variable; u(k-1)—the number of the last electrical stimulation pulse; F(x(k-1),u(k-1))—the measured value of the last state variable Non-linear mapping of the value and the number of electrical stimulation pulses;

h(k+j|k)—电刺激前向脉冲数的多项式组合；h(k)—电刺激脉冲宽度的控制量调节幅度；δ_j—控制区间的优化系数； h(k+j|k)—the polynomial combination of the number of forward pulses of electrical stimulation; h(k)—the adjustment range of the control amount of electrical stimulation pulse width; δ _j —the optimization coefficient of the control interval;

Claims

1. a functional electric stimulation closed-loop control system for electromyographic signal feedback, it is characterized in that, described system comprises:

Functional electrical stimulato, adopts the muscle of pulse current stimulating controlled device;

Electromyographic signal acquisition device, the original electromyographic signal produced when catching electricity irritation muscle, and pretreatment is carried out to original electromyographic signal, extract the absolute average amplitude of electric stimulation pulse number, electronic pulse width and myoelectricity;

A muscle contracting model, adopts the Hammerstein model containing time lag item to carry out modeling to electric stimulation pulse number and the absolute average amplitude of myoelectricity, adopts the mode of Kalman filtering to carry out parameter identification to muscle model;

Model predictive controller, according to reference to muscle excitation track and pattern and parameter identification result, by the controlled quentity controlled variable of the electronic pulse width of the optimum of optimal preventive control algorithm determination pulse current, again this controlled quentity controlled variable is fed back to functional electrical stimulato, parameter renewal is carried out to functional electrical stimulato, make the pulse current needed for functional electrical stimulato output, and real-time Self Adaptive Control is carried out to functional electrical stimulato.

2. the functional electric stimulation closed-loop control system of a kind of electromyographic signal feedback according to claim 1, is characterized in that: described original electromyographic signal is compound muscle action potential CMAP, i.e. M ripple.

3. a functional electric stimulation closed loop control method for electromyographic signal feedback, it is characterized in that, described method comprises multiple step:

S1: functional electrical stimulato initialize and signal synchronous;

S2: the original electromyographic signal produced when electromyographic signal acquisition device gathers electricity irritation muscle, and pretreatment is carried out to original electromyographic signal, extract the absolute average amplitude of electric stimulation pulse number, electronic pulse width and myoelectricity;

S3: adopt the Hammerstein model containing time lag item to carry out modeling to electric stimulation pulse number and the absolute average amplitude of myoelectricity, adopts the mode of Kalman filtering to carry out parameter identification to muscle model, obtains the two-part coefficient of linear processes of model;

S4: according to reference to muscle excitation track and pattern and parameter identification result, by the controlled quentity controlled variable of the electronic pulse width of the optimum of optimal preventive control algorithm determination pulse current, again this controlled quentity controlled variable is fed back to functional electrical stimulato, parameter renewal is carried out to functional electrical stimulato, make the pulse current needed for functional electrical stimulato output, and real-time Self Adaptive Control is carried out to functional electrical stimulato.

4. the functional electric stimulation closed loop control method of a kind of electromyographic signal feedback according to claim 3, is characterized in that: the pretreatment described in step S2 comprises:

1. amount of distortion process is gone: within an electricity irritation stimulation cycle, set threshold value masterplate do amplitude features;

2. absolute value and the meansigma methods of original electromyographic signal is extracted; Obtain the electromyographic signal amplitude that can be easy to control and identification;

3. calculating mean value needs window width: window width is the approximate integral value that rounds up of the sample frequency of electromyographic signal and the ratio of electric stimulation pulse power frequency;

4. normalized: the stability ensureing Kalman filtering.

5. the functional electric stimulation closed loop control method of a kind of electromyographic signal feedback according to claim 3, is characterized in that: the computing formula of parameter identification described in step S3 is:

In formula, the absolute average of y (k)-original electromyographic signal; The absolute average of electromyographic signal during y (k-i)-backward;

The electronic pulse width of u (k)-original electromyographic signal; u ^j(k-i) the j power of-backward electronic pulse width;

A _ithe controlled excitation system linear term parameter of (k)-to be identified; b _ithe controlled excitation system nonlinear terms parameter of (k)-to be identified; c _jthe controlled excitation system nonlinear terms parameter of (k)-to be identified;

K-electricity irritation iterative cycles number; I-linear term kinetic order; L-linear term exponent number upper limit; J-power item kinetic order; M-nonlinear terms exponent number upper limit; N-nonlinear terms exponent number upper limit;

-model linear autoregression part; -model nonlinear autoregression part.

6. the functional electric stimulation closed loop control method of a kind of electromyographic signal feedback according to claim 3, it is characterized in that: the muscle model in step S3 is converted to state space construction, the form of described state space construction is:

x(k)＝A(k)x(k-1)+B(k)Φ(u(k-1))

y(k)＝x ₁(k)

In formula, the absolute average of y (k)-original electromyographic signal;

X (k)-state equation variable; X (k-1)-laststate equation variable; x ₁(k)-laststate equation variable header element;

A (k)-coefficient matrix; B (k)-coefficient matrix;

Φ (u (k-1))-about the non-linear matrix of electric stimulation pulse number; U (k-1)-upper electric stimulation pulse number;

K-electricity irritation iterative cycles number.

7. the functional electric stimulation closed loop control method of a kind of electromyographic signal feedback according to claim 3, it is characterized in that: described parameter identification comprises first verifies identification process, adopt the kalman filter method based on forgetting factor to carry out parameter identification, it first verifies that the formula of identification process is:

In formula, -prior blur identification state matrix; -priori covariance matrix; -PREDICTIVE CONTROL exports;

X (k-1)-laststate variable measurements; U (k-1)-upper electric stimulation pulse number; The nonlinear mapping of F (x (k-1), u (k-1))-laststate variable measurements and electric stimulation pulse number;

A (k-1)-coefficient matrix to be identified; A ^t(k-1)-coefficient matrix transposition; U (k-1)-upper electric stimulation pulse number;

-prior blur identification state matrix header element; K-electricity irritation iterative cycles number; T-matrix transpose operation; λ-forgetting factor.

8. the functional electric stimulation closed loop control method of a kind of electromyographic signal feedback according to claim 3, is characterized in that: the parameter described in step S4 upgrades and comprises rear checking renewal process, verifies that thereafter the formula of renewal process is:

S (k)-Kalman filtering coefficient matrix; H (k)-Kalman filtering coefficient matrix; H ^tthe transposition of (k)-Kalman filtering coefficient matrix;

K (k)-state updating iteration coefficient matrix; S ^-1k ()-Kalman filtering coefficient matrix is inverse;

X (k)-state variable; The absolute average of y (k)-original electromyographic signal;

P (k)-measured value coefficient matrix; -estimate measured value coefficient matrix;

K-electricity irritation iterative cycles number; T-matrix transpose operation; I-unit matrix; λ-forgetting factor.

9. the functional electric stimulation closed loop control method of a kind of electromyographic signal feedback according to claim 3, it is characterized in that: encourage track and pattern according to reference to muscle, PREDICTIVE CONTROL is carried out for muscle model, choose following cost function, and by this cost function, PREDICTIVE CONTROL carried out to muscle model:

in formula, J (k)-cost function; The forward prediction of j|k-electricity irritation iterative cycles or control kinetic order; -forward prediction myoelectricity excitation amplitude;

U _d(k+j)-expect muscle excitation control objectives; ε _jthe optimized coefficients of-forecast interval;

The Polynomial combination of h (k+j|k)-electricity irritation forward pulse number; The controlled quentity controlled variable amplitude of accommodation of h (k)-electric stimulation pulse width; δ _jthe optimized coefficients of-control interval;

Np-preset forecast interval; Nu-default control is interval;

K-electricity irritation iterative cycles number; J-exponent number dynamic value; D-expected value subscript.