A new dataset for the detection of hand movements based on the SEMG signal

Abstract

in this article, we would like to present a new dataset (DS-dataset) designed to detect hand movements based on SEMG (surface electromyography) signal. This DS includes data from 42 healthy people and seven hand movements, which included three complete arm movements, i.e. punch, grip, finger touch, open hand, three-finger movements, i.e. flexion of the index finger, flexion of the middle finger, flexion of the ring finger, and one waiting state. This data was obtained using BTS's state-of-the-art Free-EMG 10-channel recorder. Based on the data in DS, the characteristic vector of the signal was generated, and were classified using classical classification algorithms (support vector machine - SVM, random forest - RF and k-nearest neighbor algorithm - k-NN). The presented DS can be used as a basis for determining the localization of electrodes and for detecting hand movements when receiving the SEMG correctly.