基于递归量化分析的表面肌电特征提取和分类

Recurrence quantification analysis based on surface EMG signal feature extraction and classification

利用展拳、握拳和腕屈、腕伸时从前臂分别检测的两路表面肌电(surface electromyography,SEMG)信号,对四种动作进行了分类研究.先采用移动平均法(moving average,MA)和一阶差分法确定SEMG信号中对应的每个动作波形的起止点,再利用递归量化分析(recurrence quantifica-tion analysis,RQA)方法提取各种动作波形的非线性特征参量(确定率、递归率等),由两路SEMG信号的这些特征参量构成特征矢量,输入BP(back propagation)神经网络,完成对不同动作的分类.研究结果表明,将利用递归量化分析得到SEMG信号的几种非线性参量作为特征值,对不同动作进行分类能够获得较高的分类准确率.

A new method for surface electromyography （SEMG） signal feature extraction based on recurrence quantification analysis （RQA） was proposed. The classification performance of four types of forearm movement using two channel SEMG signals from the forearm was researched. The classical moving average technique and first order differential were used for segmentation. Recurrence quantification analysis was adopted as an effective feature extraction technique while artificial neural networks were used for classification. Results of classifying four types of forearm movement signals gave a higher identification rate.