Abstract
Electromyography (EMG) consists of the measurement and recording of the electrical potential generated by the activation of muscle fibers when performing voluntary or involuntary movements. Therefore, electromyography signals (EMGs) are directly linked to the movement performed by a person. Hence, the study of surface EMGs to determine the movement a person is performing to control exoskeletons for post-stroke rehabilitation has become increasingly popular in recent years; particularly towards bilateral rehabilitation (BLR). Bilateral rehabilitation provides the patient the opportunity to control the intensity and frequency of the rehabilitation exercises. This paper introduces an onset detection method and a movement identification algorithm to differentiate (identify) among five different movements of the upper limb; abduction (AB), adduction (AD), flexion of the upper limb (FUL), extension of the upper limb (EUL) and AB followed by arm to the front (ABF). The movement identification algorithm focuses on the activation of muscle fibers within a single muscle when performing different movements; rather than a comparison between flexor and extensor muscles. This algorithm was evaluated using surface EMG recordings measured on healthy subjects at the Deltoid Muscle. Prior to the movement identification, the proper EMG preprocessing, feature extraction and onset detection of each EMGs recording was performed using Matlab. Two features were extracted from each channel, the root mean square (RMS) and the contractility characteristic of the muscle. An algorithm is then followed to identify the movement performed by a person. Results have shown a highly percentage of accuracy for both, the onset detection and movement identification algorithms. © 2014 Springer-Verlag Berlin Heidelberg.