摘要
针对上肢截肢患者的康复训练,结合上肢残肢肌电信号,获取并识别截肢上肢动作。从上肢表面肌电信号中提取短时能量和短时过零率特征,实现对动作段的分割,构建特征向量,使用SVM分类器对多样本数据进行统计分类,生成最优分类面。实验中,针对伸屈腕动作信号中的连续动作段,该方法总体识别率为99.1%,实验结果表明,肌电信号的短时特征基于时域累积,可以对上肢肌电信号动作进行鲁棒识别,是截肢上肢康复训练系统的有效方法。
Aiming to recognize the upper limb amputation electromyographic signal, the short time energy and zero feature of the signal were presented. The edge of the motion section from the signal was detected and the motion feature vectors were construc- ted based on motion segment. Support vector machine training was presented and an optimal classification plane was generated. The classifier can get a recognition ratio of 99.1% in the test database and it can be used to recognize the motion of the electro- myographic signal.
出处
《计算机工程与设计》
北大核心
2015年第12期3344-3348,共5页
Computer Engineering and Design
基金
湖北省自然科学基金项目(2014CFB786)
湖北省高等学校青年教师深入企业行动计划基金项目(XD2014146)
湖北省高等学校优秀中青年科技创新团队计划基金项目(T201206)
武汉工程大学科学研究基金项目
关键词
人机交互
上肢截肢
肌电信号
支持向量机
虚拟现实
动作识别
human computer interactive
upper limb amputation
electromyographic signal
support vector machine
virtual reality
motion recognition
