Finger Movement Classification from Myoelectric Signals Using Convolutional Neural Networks

Abstract

Human arms possess a number of degrees of freedom and a complicated control system. In spite of its complex design, the user efforts to perform various actions are quite minimal. Robotic hands are a mere counterpart of the natural hand, with similar flexion and extension capabilities. Over the past decades, the robotic arms play a significant role in certain applications like aiding the patients and doctors. Surface EMG signals can be retrieved in a non-invasive manner for predicting the movements of the person for the simultaneous control of the robotic arm. With Deep Learning presently in the field of Artificial Intelligence, it has paved the way for a robust and accurate system. In this paper, a Deep learning based classification technique is adopted for the classification of EMG signals collected using Myo band for the flexion of various fingers at resting position. A Convolutional Neural Network based model is presented in the paper to classify 5 different actions such as thumb flexion, index flexion, ring flexion, little flexion and rest position from the data collected from 10 subjects. An accuracy of 72.5% is observed over the custom dataset after filtering out the raw data using signal processing techniques. The application appears to be a promise for the medical industry as the proposed model can be incorporated for the control of a robotic arm for performing various operations. Results are provided to support the problem statement.