基于多尺度模糊熵的动作表面肌电信号模式识别  

Pattern Recognition of Surface Electromyography Signal Based on Multi-scale Fuzzy Entropy

作　　者：邹晓阳[1] 雷敏[1] 

机构地区：[1]上海交通大学振动冲击噪声研究所,上海200240

出　　处：《生物医学工程学杂志》2012年第6期1184-1188,共5页Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(10872125);机械系统与振动国家重点实验室基金资助项目(MSV-MS-2010-08);教育部留学回国人员科研启动基金资助项目;高等学校学科创新引智计划项目(B06012)

摘　　要：动作表面肌电(SEMG)信号是一种从皮肤表面采集的复杂电信号,它的模式识别在人体假肢和人—计算机交互系统等实际应用中非常重要。为了提高识别率,提出一种将模糊熵(FuzzyEn)和多尺度分析相结合的方法。该方法从动作SEMG信号非线性和非平稳特性的角度出发,引入了多尺度模糊熵(MSFuzzyEn)特征,并应用到人体前臂六类动作SEMG信号的模式识别中。首先利用小波分解对原始信号进行多尺度分解。然后计算MSFuzzyEn并将其作为特征向量输入支持向量机(SVM)进行识别,平均识别率达到97%,比利用原始信号的FuzzyEn进行识别时提高3%。结果表明,利用MSFuzzyEn对动作SEMG信号进行模式识别效果良好。Action surface electromyography （SEMG） signals can be acquired from human skin surface. Its pattern recognition plays a very important role in practical applications such as human prosthesis and humancomputer inter face systems. For the purpose of increasing the recognition accuracy, we proposed a new recognition method combi ning fuzzy entropy （FuzzyEn） with multiscale analysis. Considering the nonlinear and nonstationary characteristics of the SEMG, a multiscale fuzzy entropy （MSFuzzyEn） feature was introduced and applied to the pattern recogni tion of six type action SEMG signals of the forearm. Firstly, multiscale decomposition was applied to original signal using wavelet decomposition. Then MSFuzzyEn of the decomposed signals were calculated and inputted to support vector machine （SVM） for classification as feature vectors. The mean recognition accuracy reached 97%, which was 3% greater than that when FuzzyEn of original signal is applied to the classification of SEMG signals. The results have proved that the MSFuzzyEn is effective and precise in the classification of action SEMG signals.

关 键 词：表面肌电信号 模糊熵 小波分解 支持向量机 模式识别 

分 类 号：TN911.7[电子电信—通信与信息系统]