On optimal and varying decompositions for transradial contraction force prediction in upper-limb prosthesis

Abstract

• Enhanced means of predicting phantom motion using a novel signal decomposition method. • A varying gain control system for an adaptive means towards dealing with force contraction variation. • Adaptive Pattern Recognition. The abilities to recognise an intended gesture alongside its associated contraction force are both useful features which could contribute towards the advancement of bionic upper-limb prosthesis. A substantial amount of related literature appears to address these problems separately instead of as part of a sequential control system. As part of an ongoing research on this topic, this work addresses the second portion of a multi-stage advanced prosthesis control system on the prediction of the contraction force used in producing a specific gesture motion. This problem was tackled using a novel decomposition method whose decomposition parameters could be varied based on a recognised gesture motion for an optimal decomposition of the sources signal, which is capable of maximising the prediction accuracy of the control system. The results showed a 10–20% increase in the classification accuracy using this method when compared with processing done with purely the raw acquired signal. Subsequent work would now involve the testing of this proposed control system on other categories of amputees such as transhumeral and shoulder disarticulation, in order to investigate its generalisation capability.