Functional Separation of EMG Signals via ARMA Identification Methods for Prosthesis Control Purposes

TLDR

Multifunctional control of artificial limbs via EMG requires reliable recognition of distinct functions from recorded signals while also meeting constraints of weight, cost, and computation time. The study proposes a fast parametric‑recognition algorithm that identifies ARMA and Kalman filter parameters from EMG time series to solve the recognition problem. The approach derives ARMA and Kalman filter parameters from EMG time series using a rapid parametric‑recognition algorithm. The identified parameters enable discrimination among a few upper‑extremity functions, and the study discusses practical control and hardware issues to demonstrate the approach’s validity.

Abstract

Multifunctional control of artificial limbs via electromyographic (EMG) actuation requires means for reliably recognizing or distinguishing between the various functions on the basis of the recorded EMG data. Furthermore, constraints of weight, cost, and computation time on practical prosthesis application must be satisfied. An approach to the aforementioned recognition problem is given in terms of deriving a fast parametric-recognition algorithm whereby the autoregressive-moving-average (ARMA) parameters and the Kalman filter parameters of the EMG time series are identified. It is shown that the resulting identified parameters yield sufficient information to discriminate between a small number of upper extremity functions. Problems involved in practical prosthesis control via the present approach and problems of hardware realization are discussed to illustrate the validity of the approach.