A neuromuscular interface for the elbow joint

Abstract

The increasing popularity of using biosignal interfacing with assistive devices and users who are physically disabled sees research being split into disparate areas of electromyography (EMG) signal filtering, feature extraction and interpretation, and specific areas of control. This paper presents the development of a neuromuscular interface (NI), which has the sole function of converting the EMG signals of a particular joint into a predicted torque or displacement. The proposed system consists of an analogue signal filtering PCB and microcontroller that uses a neuromusculoskeletal model of the elbow joint to predict elbow motion. The NI only relies on the EMG signal and the raw EMG is not enhanced in any way. Trials in real time have shown that after tuning with genetic algorithms and some manual adjustments to account for limitations in genetic algorithms, the interface is capable of predicting motion with an RMSE value of 13.0°. This work provides an initial platform for the development of generic NI hardware that can be applied to an unlimited number of research applications.