Design and Signal Quality Assessment of a Novel Low-Cost Wireless EMG Sensor

Abstract

Electromyography (EMG) sensors have been widely used for assessing muscle activation state and human-robot interaction (HRI). However, available commercial EMG sensors are expensive and inconvenient for real-time applications. The aim of this study is to design and validate a novel low-cost wireless EMG sensor. Specifically, the wireless EMG sensor is equipped with an analog-front-end (AFE), radio frequency (RF) module, and controller area network (CAN) bus, which achieves wireless data transmission and provides an interface for real-time HRI. The evaluation of the designed system was done by a series of exercises performed by volunteers. Raw signal., mean absolute value (MAV), zero crossings (ZC), slope sign changes (SSC), waveform length (WL), and median frequency (MDF) are used for comparing wireless EMG sensors and commercial sensors. In addition, Spearman's correlation (SC), and intra-class cross-correlation (ICC) are used as two validation indicators. The result shows that the SC and ICC between the raw signals of the two types of sensors are 0.6156 and 0.7780, respectively. The SC of MAV, ZC, SSC, WL, and MDF are 0.9708, 0.6740, 0.8023, 0.9663, and 0.9476, respectively. The ICC of MAV, ZC, SSC, WL, and MDF are 0.9670, 0.7340, 0.7910, 0.9540., and 0.8230, respectively. The results indicate a good agreement between the two systems and confirm the reliability of the wireless EMG sensor's design. The wireless EMG sensor can be further applied to real-time HRI and muscle fatigue assessment.