A Neuro-Fuzzy Inference System For Recognizing Speed And Stiffness Of Upper Limb Movements Based On SEMG

Electromyography (EMG) signals have been the subject of considerable researchefforts in recent years. They are found to show accurate patterns for different activities ofmuscles.This has carved the way for its applications in clinical diagnosis,andrehabilitation in addition to prosthetic control and of assistive devicescontrol. Correctlyidentifying these patterns helps to provide better control of assistive devices. Thisresearch work demonstrates the effectiveness of the adaptive neuro-fuzzy inferencesystem (ANFIS) on recognizing the pattern of certain muscle activities. This thesispresents an attempt of classifying EMG signals based on movement of a human elbowand forearm (both flexion/extension and probation/suspiration) for eightmovementpatternsin multiples ways based on various speeds and stiffness’s; as part ofdevelopment of signal classification. Moreover, recorded EMG signals from fivevolunteers accumulating six muscles for each individual through a variety of movements.From the acquired EMG data, statistical features are extracted and are applied as inputs tothe classifiers. With the maximum identification accuracyrate of100%and an averageclassification accuracy of98.53%, the proposed ANFIS system has beenproved to besuperior in comparisonwith relevant studies to date.