摘要
表面肌电信号(Surface electromyography,s EMG)已广泛应用于手部动作识别。为提高动作识别精度,研究者往往需要采集多个通道s EMG信号,从而增加应用复杂性,针对这一情况,提出一种基于单通道s EMG分解的手部动作识别方法。使用单通道电极采集人体上臂肌肉s EMG,将其分解为6个运动单元动作电位序列,过程包括:二阶差分滤波、阈值计算、尖峰检测、分层聚类;然后,提取绝对值积分、最大值、非零中值、半窗能量等特征,并采用主元分析法降维;最后,利用支持向量机分类识别5种不同手部动作,精度达到80.4%。而采用未融合s EMG分解的传统方法,动作识别精度仅有约70%。
Surface electromyography（s EMG） has been applied extensively in gestures recognition. In order to improve the recognition accuracy, multi-channel s EMG is conventionally sampled, which also increases the complexity of applications. To solve the problem, a novel gesture recognition method based on s EMG decomposition is proposed. Sampling s EMG signals from the muscle of human upper limb by a single-channel electrode; then decomposing the s EMG into six motor unit action potential trains（MUAPTs） and the decomposition process includes 2-order differential filtering, threshold calculation, spike detection and hierarchical clustering. Afterwards, the features, including integral of absolute value, maximum value, median of non-zero value and semi-window energy, are extracted to form a feature matrix, whose dimension is then reduced by the principal component analysis. Finally, support vector machine is employed to recognize five different hand gestures, and 80.4% of accuracy can be obtained, while only about 70% of recognition accuracy can be achieved by traditional methods without s EMG decomposition.
出处
《机械工程学报》
EI
CAS
CSCD
北大核心
2016年第7期6-13,共8页
Journal of Mechanical Engineering
基金
国家高技术研究发展计划(863计划
2015AA042302)
国家自然科学基金(61273355
61503374
61573340)
机器人学国家重点实验室自主课题(2015-z06)资助项目
关键词
表面肌电信号
运动单元动作电位序列
分层聚类
主元分析支持向量机
sEMG
motor unit action potential trains
hierarchical clustering
principal component analysis
support vector machine
