摘要
为改善足下垂患者步态,研究了踝关节不同动作的表面肌电信号特征分类.本文采集踝关节在不同动作下,对应胫骨前肌、腓肠肌、腓骨长肌和拇长伸肌的表面肌电信号,采用小波包分解方法进行肌电特征提取,获得小波包系数能量、方差统计特征量;利用支持向量机方法实现踝关节4种不同动作模式的肌电特征分类.实验结果表明,采用具有良好奇异特性的小波包能量、对数方差构成的肌电特征向量,对踝关节动作进行模式识别,其正确率远高于通过提取肌电信号时域或者频域特征进行模式分类的正确率,达到了92.8%的平均分类正确率.该特征提取方法以及支持向量机分类器,可以应用于踝关节动作识别和机器人康复工程.
In order to improve the patient＇s foot drop gait, the paper discussed the feature classification of the surface Electromyography （EMG） signal to the different motions of the ankle joint. First, the surface EMG signals of anterior tibial muscle, gastrocnemius muscle, peroneus longus and extensor hallu- cis longus under the different motions of the ankle joint are collected. Then,wavelet packet decomposi- tion method was utilized to extract sEMG feature. Next, the method of support vector machine （SVM） was used to classify four different motion patterns of the ankle joint. Experimental results shown that u- sing the wavelet packet coefficient energy and Log variance features with good singularity as feature vec- tor, ankle joint movement was recognized,which average correct rate of SVM classifier achieves 92. 8 %to ankle joint action, is much higher than adopting only time domain or frequency domain feature extrac- tion method. This proposed feature extraction method and support vector machine classifier can be effec- tively applied to the motion recognition of the ankle joint and the robot rehabilitation project.
作者
胡文龙
乔晓艳
HU Wenlong QIAO Xiaoyan(College of Physics and Electronics Engineering, Shanxi University, Taiyuan 030006, Chin)
出处
《测试技术学报》
2017年第2期100-106,共7页
Journal of Test and Measurement Technology
基金
山西省回国留学人员科研资助项目(2014-010)
关键词
踝关节
表面肌电信号
小波包统计特征
支持向量机
ankle joint
surface EMG
wavelet packet statistic feature
support vector machine
