摘要
针对表面肌电信号的非线性和非平稳性等特点,提出了一种主元分析与核LDA判别分析相结合的表面肌电信号特征识别新方法;通过虚拟仪器采集桡侧腕屈肌和肱桡肌两路表面肌电信号,并提取其特征参数平均绝对值和均方根,采用主元分析法对表面肌电信号特征参数进行压缩降维,应用核LDA判别分析法对降维后的数据进行分类识别;经过实验表明,该方法将表面肌电信号的特征参数由4维降到2维,减小了数据的冗余度,能够成功的从表面肌电信号中识别握拳、展拳、手腕内翻和手腕外翻四种动作,识别率高达96%。
For the nonlinear and non--stationary characteristics of SEMG, a new SEMG feature recognition method which includes prin- cipal component analysis and kernel linear discriminant analysis is proposed. Through the acquisition of two channels of SEMG on flexor carpiradialis and brachioradialis with virtual instrument, working out the mean absolute value （MAV） and root--mean--square （RMS） as feature parameters, the principal component analysis （PCA） method is used to compress SEMG feature parameters dimension reduction, and kernel discriminant analysis method （KPCA） is adopted to classify the data after dimension reduction. The experiments show that it reduces the characteristic parameters of surface EMG signal consist from 4 dimensions to 2 dimensions and data redundancy, and successfully identi- fies four kinds of motions, such as Making a fist, fist exhibition, varus wrist and wrist valgus with recognition rate up to 96%.
出处
《计算机测量与控制》
北大核心
2014年第2期575-577,共3页
Computer Measurement &Control
基金
国家自然科学基金项目(51205372)
