基于ART2神经网络的手势动作肌电信号识别  

Hand gesture SEMG signal recognition based on ART2 neutral network

作　　者：李媛媛[1] 陈香[1] 张旭[1] 杨基海[1] 

机构地区：[1]中国科学技术大学电子科学与技术系,安徽合肥220027

出　　处：《中国科学技术大学学报》2010年第8期829-834,共6页JUSTC

基　　金：国家自然科学基金(60703069);中国高技术研究发展(863)计划(2009AA01Z322);中国科学技术大学创新团队培育基金资助

摘　　要：针对手势动作肌电信号识别中存在的识别效率和稳定性问题,采用一种基于ART2神经网络的手势识别方法,并进行了单用户和多用户的实验研究.对8名受试者、8类手势动作模式的单用户实验,取得了较高的识别正确率,与BP网络相比,ART2分类器具有识别率高、实时性好、鲁棒性强的优点;同时,多用户的实验结果表明,ART2网络对手势动作肌电信号的识别具有良好的自适应性和稳定的分类能力.In view of the problems of efficiency and stability existing in hand gesture SEMG signal recognition, a method of hand gesture recognition based on the adaptive resonance neutral network 2 （ART2） was proposed. Compared with the back propagation （BP） network classifier, ART2 classifier effectively improves the recognition rate, reduces the signal processing time, and is not sensitive to individual differences in the single-user experiments conducted on eight classes of hand gestures for eight subjects. Multi-user experimental results also indicate that ART2 classifier has good adaptive capacity and stability in discriminating hand gesture SEMG signals.

关 键 词：表面肌电信号 手势识别 自适应共振 ART2神经网络 

分 类 号：R318.0[医药卫生—生物医学工程]