文献类型：期刊文献

中文题名：基于肌音信号的头部动作模式识别

英文题名：Pattern Recognition of Head Movement Based on Mechanomyographic Signal

作者：顾晓琳[1];吴清[1];夏春明[1];章悦[1];钟豪[1]

机构：[1]华东理工大学机械与动力工程学院,上海200237

年份：2017

卷号：43

期号：5

起止页码：704

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：北大核心:【北大核心2014】；CSCD:【CSCD_E2017_2018】；

语种：中文

中文关键词：肌音;头部动作;特征提取;小波包;双谱

外文关键词：mechanomyography; head movement; feature extraction; wavelet packet; bispectrum

摘要：肌音信号(MMG)是一种肌肉收缩时发出的低频信号,通过测量分析颈部前后两侧的胸锁乳突肌和头夹肌的肌音信号,成功识别点头、抬头、左摆、右摆、左转、右转6个头部动作模式。实验中采集了4个通道的数据,经滤波、归一化的预处理后,用不等长分割法分割出动作帧。提取了动作帧的小波包系数能量及双谱对角切片特征,经主元分析法(PCA)和Fisher线性判别分析(FLDA)降维,用支持向量机(SVM)分类。最后对小波包系数能量和双谱对角切片特征进行FLDA降维,识别率达95.92%。
Mechanomyography （MMG） is a low frequency signal when muscle is contracted. Four channel MMG signals are collected from the sternocleidomastoid （SCM） muscles and splenius capitis （SPL） muscles in the subjects’ neck when they bowed head, raisedhead, bent side to le ft, bent side toright, turned to le ft, and turned to right, i. e. , six action modes, which could be successfully recognized. Thefour channel MMG signals were then filtered, normalized, and divided using unequal length segmentation algorithm. After extracting the energy features of wavelet packet coefficients and the feslices of spectrum,the dimension of features were reduced by principal component analysis （PCA ） or fisher linear discriminant analysis （FLDA）. Finally, all the features were classified by SVM classifier. When the features of wavelet packet coefficients energy and diagonal slices of spectrum went through FLDA dimension reduction,the recognition ratewere up to 95. 92%.