一种适用于手势动作sEMG信号识别的改进型模糊推理分类器  

A modified fuzzy inference classifier suitable for pattern recognition of hand gesture sEMG signals

作　　者：涂有强[1] 陈香[1] 张旭[1] 赵章琰[1] 杨基海[1] 

机构地区：[1]中国科技大学电子科学与技术系,合肥230027

出　　处：《北京生物医学工程》2008年第4期362-366,392,共6页Beijing Biomedical Engineering

基　　金：国家自然科学基金(60703069)资助

摘　　要：提出了一种基于自适应提取模糊规则的改进型模糊推理分类器,其中,模糊规则的提取采用由势函数法初始化聚类中心的K-means聚类算法,分类器的训练采用基于梯度下降算法的最小均方误差准则来实现。此改进型模糊分类器克服了基于K-means聚类算法提取模糊规则的模糊推理分类器需要手工设定模糊规则数目和对初始化参数非常敏感的两大缺点。对10位受试者的6类手势动作sEMG信号的分类研究结果表明,此改进型模糊推理分类器的分类能力优于未改进的模糊推理分类器,且具有效果稳定、自适应提取模糊规则、对初始化参数不敏感以及可排除孤立点的影响等优点。In this article,a modified fuzzy inference classifier based on adaptive fuzzy rules extraction was proposed for the pattern recognition of hand gestures sEMG signals. An adaptive algorithm,which initializes the clustering centers of the K-means clustering arithmetic using mountain method, was adopted to extract fuzzy rules. Gradient descent algorithm based on mean square error function was used in the training procedure of the classifier. For exploring the capability of this modified fuzzy inference classifier, pattern recognition of six classes of hand gesture sEMG signals form 10 subjects had been implemented. Experimental results demonstrated that,compared with the un-modified fuzzy inference classifier based on Kmeans fuzzy rules extraction, the proposed classifier got better recognition accuracies. The stability,adaptive capacity,less sensitivity to initial parameters and the ability to eliminate the bad effect of some isolated points make this modified fuzzy inference classifier is powerful in discriminating hand gesture sEMG signals.

关 键 词：手势动作sEMG 模糊分类 势函数法 K-MEANS聚类算法 梯度下降算法 

分 类 号：R318.04[医药卫生—生物医学工程]