表面肌电分类研究及基于dsp的算法实现（可复制）.pdf

上海交通大学硕士学位论文摘要

表面肌电分类研究及基于 DSP 的算法实现
摘要
表面肌电信号的分类与识别是近年来国内外研究非常广泛的一个
课题。如何从表面肌电信号中有效地提取信息并进行正确的动作识别
是实现假肢控制的重要内容。本文主要以小波变换和决策树为理论基
础,对前臂肌肉在不同运动模式下的表面肌电信号进行深入研究。
本文创新性地提出了采用模糊小波包方法,即将小波包能量结合
信息判别度来提取表面肌电信号的特征。实验表明,与其他利用小波
理论进行表面肌电信号特征提取的方法相比,该方法提取的特征更加
具有区分识别能力。在对已提取的特征进行分类的过程中,本文独创
性地采用了决策树进行分析和判断,并和其他传统方法综合比较。实
验表明,将模糊小波包和决策树综合应用于表面肌电信号的分类识别
可以达到较高的识别效果。
本文采用现今最流行的 DSP 平台实现表面肌电信号的分类识别算
法,研究和论证算法在 DSP 上快速实现的可能性,探讨算法应用于
假肢控制的可行性,这对表面肌电信号研究的发展和应用有一定的实
际意义。

关键词:表面肌电信号,小波变换,决策树,模糊小波包,小波
包能量,信息判别度,DSP

I
上海交通大学硕士学位论文 ABSTRACT

RESEARCH ON CLASSIFICATION OF SEMG AND
ALGORITHMIC REALIZATION
BASED ON DSP
ABSTRACT
In the last few years classification and recognition of sEMG has been
widely researched. Effective feature extraction and accurate identification
is the crucial content involved in realizing prosthesis control. On the basis
of wavelet transform and decision tree, this dissertation deeply discussed
sEMG in different movement patterns.
This dissertation creatively proposed a fuzzy wavelet packet based
feature extraction method, which bined by wavelet packet energy
and information criterion, to extract sEMG features. Compared with other
methods based on wavelet transform, this method can extract features
which are easier to be distinguished. In the process of classification, this
dissertation chooses decision tree to analyze and estimate the features.
Finally the experiments prove that this method has a high correct
classification rate.
This dissertation realizes the classification and recognition algorithm
by using the most popular DSP platform, researches the possibility of
algorithmic realization on DSP, and discusses the feasibility of
II
上海交通大学硕士学位论文 ABSTRACT
algorithmic application to prosthesis control, which has a practical
significance to development and application of research on sEMG.


KEY WORDS: sEMG, wavelet transform, decision tree, fuzzy wavelet
packet