3、ntControlandRobotics,HangzhouDianziUniversity,Hangzhou310018,China)Abstract:Action pattern recognitionoflimbsusingsEMGisthebasis for bioniccontrolofaprosthetichand. Inconsiderationofthegeneration mechanismofsEMG,theapproximate entropyandthefractaldimension,which featurethesEMG’smorphologic
4、alcharacteristicsincludingcomplexityandoverallself-similarity, waschosenasthe featurevector ofpatternrecognitiontoimproveactionmoderecognitionrate. Inthemeantime,a K Nearest Neighbor (KNN)modelincrementallearningmethod with incrementallearningability,was presented as a classifierofpatte
5、rnrecognition.Inpatternrecognitionexperimentofclassifyingfourfine movements ofthewrist(namelywristextension,wristflexion,wristpronation,wristsupination)with10participants,thecorrectmoderecognitionrateisabove92.5%. In a contrastexperimentthat wasdesignedtoevaluatethe effectsof theincrement