Biomimetic finger control by filtering of distributed forelimb pressures

A linear filter was developed for decoding finger commands from volitional pressures distributed within the residual forelimb. Filter parameters were based on dynamic pressures recorded from the residual limb within its socket, during specific finger commands. A matrix of signal features was derived from eight-dimensional pressure vectors, and its pseudoinverse comprised the filter parameters. Results with amputees showed that the filter could discriminate specific finger flexion commands, suggesting that pressure vector decoding can provide them with biomimetic finger control.

I. Introduction

While prosthetic terminal devices can be controlled with voluntary electromyographic (EMG) signals from the residual limb, their function is generally limited to one degree Fig.  1. Discriminatory filtering of distributed interfacial pressure data. Filtering for fingers is shown. Filter vectors , comprised of custom weights, are applied to the signal vector, , to obtain a control vector, , an approximation of the intended movements. of freedom (DOF), i.e., grasping. A major reason for this limitation is the relative unsuitability of the EMG for joint control, since it consists of noise with little direct relationship to joint movements. Nevertheless advanced signal processing techniques continue to improve EMG-based control of prosthetic hands [1]–[5].