摘要
采用支持向量机作为分类器,通过在健康受试者前臂处安放8个表面肌肤干电极提取肌电信号,使用信号均值作为特征,以较高成功率实现人手10种姿态的分类.分类结果加窗后输出至3自由度假手控制器,实现"姿态跟随"以及"位置/力矩迭加"两种控制方法.试验结果表明,手部姿态的多模式识别使得多自由度肌电假手的控制更加柔顺,体现了较高的灵巧性与功能性.
Based on the pattern recognition method of supprt vector machine, 10 mode hand gestures have been succeccfully classified by using the average features of electromyograph (EMG) signal extracted from the healthy body's forearm through 8 dry electrodes. By feeding the windowed classifying results into the prosthetic hand's controller, two control methods, "State Following" and "Position/Force Overlaping", are implemented. Experimental results show that the prosthetic hand's control becomes more facile through classifying multi-mode hand gestures, which presents a high dexterity and functionality.
出处
《江苏大学学报(自然科学版)》
EI
CAS
北大核心
2009年第1期5-9,共5页
Journal of Jiangsu University:Natural Science Edition
基金
国家自然科学基金资助项目(50435040)
关键词
假手
肌电控制
模式识别
支持向量机
prosthetics
myoelectric control
pattern recognition
support vector machine