状态空间EMG模型：连续关节运动估计

88 浏览量更新于2024-08-27 1 收藏 1.65MB PDF 举报

"用于连续关节运动估计的状态空间EMG模型是一种创新的方法，它结合了肌电图（EMG）信号处理与人类关节运动的前向动力学，以实现对人体肢体连续运动的精确估计。传统的基于Hill的肌肉模型只能开环估计关节扭矩，而该模型扩展了这一概念，直接估计关节运动状态。通过构建状态空间模型，利用Kalman滤波等闭环预测校正技术，可以从EMG信号中排除模型不确定性及外部干扰，从而更准确地估算连续的关节运动。此外，文章还提出了一种EMG信号的归一化新方法，以减少运动估计对变化外部负载的敏感性。实验证明，这种方法在人体肘关节的实验中表现出色，与惯性测量单元的测量结果进行了比较，证实了其改进效果。" 这篇研究的核心在于构建一个状态空间EMG模型，它涵盖了以下几个关键知识点： 1. **状态空间模型**：这是一种数学模型，用于描述系统的动态行为，其中系统的状态随着时间变化，并且可以通过一组方程来表示。在本文中，状态空间模型用于将EMG信号转化为关节运动的连续估计。 2. **肌电图（EMG）**：EMG是记录肌肉活动的一种生理信号，通过分析这些信号可以推断出肌肉的收缩状态和运动信息。在这个模型中，EMG被用作输入，用于估计关节运动。 3. **Hill-based肌肉模型（HMM）**：这是一种模拟肌肉力学特性的模型，通常用于计算肌肉产生的扭矩。然而，HMM在原形态下不考虑关节运动的前向动力学，只做开环估计。 4. **前向动力学**：这是描述物体运动的物理规律，包括力、速度和加速度之间的关系。在本文中，前向动力学被整合到HMM中，使得模型能够考虑关节运动的实际动态过程。 5. **闭环估计**：与开环估计不同，闭环估计系统能够根据反馈信息不断调整其输出，以减少误差。文中采用的Kalman滤波器就是一种典型的闭环预测校正方法，它能有效地处理模型的不确定性和外部干扰。 6. **Kalman型算法**：这是一类用于线性高斯系统的最优估计方法，能够在线性化模型和噪声统计已知的情况下，提供最小均方误差的估计。在状态空间EMG模型中，Kalman滤波器用于从EMG信号中提取连续关节运动信息。 7. **EMG信号归一化**：为了提高运动估计的稳定性和准确性，文章提出了新的归一化方法，旨在消除EMG信号对不同外部负载的敏感性，确保运动估计的独立性。 8. **实验验证**：通过人体肘关节的实验，研究人员对比了基于EMG的运动估计与惯性测量单元（IMU）的数据，以此证明所提方法的有效性和改进程度。这项研究不仅提供了连续关节运动估计的新方法，还为生物力学和康复工程领域带来了重要的进展，为未来基于EMG的运动控制和分析提供了理论和技术支持。

IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 62, NO. 7, JULY 2015 4267

A State-Space EMG Model for the Estimation

of Continuous Joint Movements

Jianda Han, Member, IEEE, Qichuan Ding, Anbin Xiong, and Xingang Zhao, Member, IEEE

Abstract—A state-space electromyography (EMG) model

is developed for continuous motion estimation of human

limb in this paper. While the general Hill-based muscle

model (HMM) estimates only joint torque from EMG signals

in an “open-loop” form, we integrate the forward dynamics

of human joint movement into the HMM, and such an

extended HMM can be used to estimate the joint motion

states directly. EMG features are developed to construct

measurement equations for the extended HMM to form a

state-space model. With the state-space HMM, a normal

closed-loop prediction–correction approach such as the

Kalman-type algorithm can be used to estimate the continu-

ous joint movement from EMG signals, where the measure-

ment equation is used to reject model uncer tainties and

external disturbances. Moreover, we propose a new nor-

malization approach for EMG signals for the purpose of re-

jecting the dependence of the motion estimation on varying

external loads. Comprehensive experiments are conducted

on the human elbow joint, and the improvements of the

proposed methods are veriﬁed by the comparison of the

EMG-based estimation and the inertial measurement unit

measurements.

Index Terms—Closed-loop estimation, continuous joint

motion, electromyography (EMG), muscle model.

I. INTRODUCTION

SING ELECTROMYOGRAPHY (EMG) as control sig-

nals to realize a “friendly” human–robot interface (HRI)

has been there for assistive robots [1]–[3]. Classiﬁcation, which

distinguishes different patterns of motion from EMG, is one

of the key techniques for such an HRI [4]. The classiﬁcation

involves two steps: 1) extracting feature sets from EMG and

2) classifying different motions based on the selected feature

sets [5]. For step 1, the EMG feature sets include EMG am-

plitude, autoregressive coefﬁcients, waveform length, cepstrum

coefﬁcients, and the wavelet packet transform [6]. Certain

feature sets, such as the wavelet packet transform, have to

be used in conjunction with dimension reduction algorithms

such as principal components analysis [7] or linear discriminant

Manuscript received February 15, 2014; revised June 11, 2014 and

October 4, 2014; accepted November 7, 2014. Date of publication

January 1, 2015; date of current version May 15, 2015. This work was

supported in part by the National Natural Science Foundation of China

under Grant 61273355, Grant 61273356, and Grant 61035005.

J. Han and X. Zhao are with the State Key Laboratory of Robotics,

Shenyang Institute of Automation, Chinese Academy of Sciences,

Shenyang 110016, China (e-mail: jdhan@sia.cn).

Q. Ding and A. Xiong are with the State Key Laboratory of Robotics,

Shenyang Institute of Automation, Chinese Academy of Sciences,

Shenyang 110016, China, and also with the University of the Chinese

Academy of Sciences, Beijing 100049, China.

Color versions of one or more of the ﬁgures in this paper are available

online at http://ieeexplore.ieee.org.

Digital Object Identiﬁer 10.1109/TIE.2014.2387337

analysis [8] to yield proper signal representation. For step 2,

commonly used classiﬁcation algorithms are support vector

machine [9], artiﬁcial neural networks [10], hidden Markov

model [11], Gaussian mixture model [12], and neuro-fuzzy

algorithm [13].

The aforementioned algorithms are able to identify a limited

number of discrete motion patterns from EMG. However, re-

cently, instead of the discrete patterns, how to determine the

human’s intent of a continuous motion from EMG has become

an active issue. This is due to the requirements of several possi-

ble applications, such as powered prosthesis, exoskeletons, and

rehabilitative robots [14], [15]. In such a system, the human

motion intents have to be continuously recognized from EMG

and further reconstructed as control commands to a robotic

device, so that the robot could match the human’s intent and

then perform efﬁcient assistance.

Involving a physiological muscle model into classiﬁcation

algorithm is a way to achieve continuous EMG recognition,

and the Hill-based muscle model (HMM) is the most frequently

utilized one [16]–[19]. In [16], an EMG-based forward dynam-

ics model was proposed, which consists of muscle activation

dynamics, Hill-based muscle contraction dynamics, muscu-

loskeletal geometry, and joint forward dynamics. This model

is very complex and involves many unknown physiological

parameters, which limits its applications in real mechatronic

systems. A simpliﬁed model for controlling lower extremity

exoskeleton was developed in [18], where a step calibration

process was constructed to optimize the estimation of unknown

parameters.

However, there exist two problems while the normal Hill-

type models are used to estimate continuous motion states of

human limbs. First, a full Hill model always involves many

physiological parameters, which are very difﬁcult to be ac-

curately identiﬁed. The complex model will also introduce

extra computational burden. Thus, a simpliﬁed model is often

used for real-time applications. However, a simpliﬁed model

inherently has modeling errors that will degrade the estimation

accuracy. Normally, the measurement equations in a state-

space model will reduce the effects of modeling errors and

uncertainties. However, the Hill-type models are in “open-loop”

structure; there are no other measurement equations that can be

used as a kind of feedback to improve the estimation accuracy.

Second, the Hill-type models are often used to directly

calculate the joint torque from EMG signals [18]–[21]. For the

applications where the continuous estimations of human joint

are required, for example, to control a robotic device assisting

the movement of human limb [22], we have to further calculate

the motion states, such as angular velocity and position, from

IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.

See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

下载后可阅读完整内容，剩余8页未读，立即下载

weixin_38663526

粉丝: 3
资源: 940

状态空间EMG模型：连续关节运动估计

EMG Feature Extraction

基于自适应模糊神经网络的下肢关节运动意图估计.pdf

用于检测和识别随机发生的异常运动的增量式EMG分类模型

opendihu:用于多尺度骨骼肌模型的仿真框架，用于计算EMG信号和肌肉收缩

EMG

基于MATLAB实现的多输入多输出系统的状态空间simulink模型，可以直接在里面修改运行获得阶跃响应+使用说明文档.rar

matlab肌电信号处理代码-DTW_ANN_EPN_Matlab:这是手势识别模型的示例，用于在Matlab中管理EMG-EPN2020数据

EMG.rar_EMG MATLAB_EMG_matlab_emg activation_emg gui_emg_activat

EMG.zip_EMG

EMG.rar_EMG

最新资源