Flexion Angle Estimation from Single Channel Forearm EMG Signals using Effective Features

Abstract

Electromyography (EMG) records the electrical activity generated by skeletal muscles, offering valuable insights into muscle function and movement. To address the complexity of EMG signals, various signal analysis methods have been developed in the time and frequency domains for engineering applications like myoelectric control of prosthetics and movement analysis. In this study, EMG signals were acquired from ten healthy volunteers in different forearm positions using a Myoware Muscle Sensor and MPU6050 board. From each EMG signal, root mean square (RMS), standard deviation (STD), and mean absolute value (MAV) were computed and selected as representative features. These features were then fed into an LDA classifier to estimate forearm flexion angles. The study aims to compare the effectiveness of features calculated from the EMG signal and those derived from its discrete wavelet decomposition. The experimental results demonstrate the proposed method's efficiency in estimating forearm flexion angles using a single channel of EMG signals, achieving an average classification accuracy of 97.50 % across four gesture classes.