CLASSIFICATION OF TRANSIENT MYOELECTRIC SIGNALS FOR THE CONTROL OF MULTI-GRASP WRIST-HAND PROSTHESIS

Abstract

Decoding the neurophysiological signal generated by voluntary arm movements is one of the major challenges in rehabilitation engineering. The most investigated approach for hand prosthesis control is the continuous pattern recognition of myoelectric signals. However, this is based on the assumption that repeated muscular contractions produce consistent patterns of steady-state myoelectric signals. Notably, it is the initial, transient, phase of such signals that was shown to contain a deterministic structure. Here we investigated if both wrist and hand intended movements could be decoded from the transient phase of the myoelectric signal. Twelve healthy individuals performed one of four grasps and of five wrist movements simultaneously (20 combinations). Albeit the performance in recognizing both movements simultaneously was poor, the offline data analysis showed the feasibility of implementing a sequential wrist-hand embedded controller based on the transient phase.