非线性PCA在表面肌电信号特征提取中的应用.pdfVIP

第 20 卷 　第 10 期 传 感 技 术 学 报 Vol . 20 　N o . 10 2007 年 10 月 CHIN ES E J OU RNAL OF S ENSORS AND AC TUA TORS OC T . 2007 Nonl inear Principal Component Analysis f or Feature Extraction of SEM G L UO Z hiz eng , Z H A O Pengf ei ( ) H angz hou D i anz i Uni vers ity I ntel l i g ent Cont rol R obot ics I ns t it ute , H ang z hou 3 100 18 , Chi na Abstract :In connection wit h t he charact er of Surf ace Elect ro myo grap hy signal ( SEM G) , a new met ho d t hat u se s nonlinear Princip al Co mp onent A naly si s (NL PCA) to ext ract f eat ure f ro m SEM G wa s p ropo sed . Aft er filt ering SEM G , it utilize s NL PCA to achieve dat a co mp re ssio n , w hich t ran sfor m s multiway SEM G to one dimen sional f eat ure dat a saying p rincip al co mponent , and t hen ,outp ut s t he ext raction in p rincip al curve . NL PCA ba sing on autoa ssociative neural net wor k s wa s utilized to ext ract ed p rincip al co mponent f ro m t woway SEM G , w hich derived f ro m ulnar ext en sor mu scle and ulnar flexor mu scle of wri st re sp ec tively . Exp eriment al re sult s showed t hat , aft er p roce ssing SEM G of four hand motion p at t er n s t hat inclu din g fi st clenching , fi st unfol ding , wri st intor sion an d wri st extortion wit h t hi s met ho d , p rincip al curve s wit h goo d charact er of cat egory divi sion were p ro duced . Accor ding to t he shap e feat ure s of p rincip al curve s , motio n of hand can be reco gnized efficient ly . Key words :