A Study of Kalman Filtering Applied to Myoelectric Signal State Tracking

Abstract

The goal of this work is to reduce the response time of pattern recognition based myoelectric prostheses without compromising stability. A Kalman filter (KF) was applied in feature-space to track class transitions and to determine when features have converged towards steady-state class. The system was tested against data collected during continuous movement where subjects transitioned between seven forearm and hand motions. For various data acquisition times, the signal-to-noise-ratio obtained from filtered and non-filtered features were compared, and the system classification accuracy and processing time were compared against state-of-the-art systems. Results show that while applying the proposed system, data acquisition time can be reduced from 100ms to 20ms without compromise to the system’s classification accuracy.