Evolving Fuzzy Models for Prosthetic Hand Myoelectric-based Control Using Weighted Recursive Least Squares Algorithm for Identification

Abstract

This paper discusses the use of the weighted Recursive Least Squares (wRLS) algorithm in the identification of a set of evolving Fuzzy Models (FMs) that characterize the nonlinear finger dynamics of the human hand in the framework of myoelectric (ME)-based control of a prosthetic hand. FM inputs are ME signals obtained from eight ME sensors, and different numbers of additional model inputs obtained from past inputs and/or outputs are considered. The FM outputs are the finger angles, i.e., the midcarpal joint angles. Evolving FMs for two out of five fingers are derived. Model performance for wRLS versus Recursive Least Squares (RLS) algorithm tested against experimental data is outlined and compared.