Walking PhaseRecognition forPeople withLowerLimbDisability

This paperpresents a totalsolution on EMG signal-based walking phaserecognition forpeople withlower limbdisability. Various environmental factors suchassensed location, walking speed, andgroundinclination aretakeninto consideration inallthephasesofsignal sensing, feature extraction, feature selection, and classification. Basedon analysis onfourteen well-known feature extraction methodsin varyingenvironmental situation, thispaperproposesa methodology forselecting a goodfeature set,and then demonstrates effectiveness oftheproposed approach withthe classification results. I.INTRODUCTION UMAN bio-signals areknowntoposses enormous potentiality tooffer meansofinteraction between human andmachine, thankstoadvancement or current technology ofmeasurement andinformation processing. Various formsofhumanbio-signal including electro- encephalogram(EEG), eletrocardiogram(ECG), electro- olfactogram(EOG), electromyogram (EMG)canbeutilized inestablishing man-machine communication channels tohelp people withdisabilities. Amongthem, we areparticularly interested inEMG signal whichhasbeenwidely studied in thefield ofrehabilitation robotics duetoitscapability of revealing workings ofhumanmotornervesystem, thus directly reflecting theintention ofperson's activity. AlongwithEMG-based recognition ofmovement ofupper body,we findthatlowerbodyEMG-recognition has occupied animportant avenue inrehabilitation robotics, since thelower part notonly sustains thewholebody, butalso takes charge ofthemovement fromoneplace toanother, which are themostfundamental activities ofhuman. During walking, muscle contraction takes place incycles anditmatches the transition ofmuscleactivity. Indeed, itistheelectrical manifestation oftheneuromuscular activation associated Thisworkissupported bytheSRC/ERCprogram ofMOST/KOSEF