Myoelectric neural networks signal analysis

Abstract

The myoelectric signal analysis strategy in such situations is to have the user/subject generate a set of repeatable muscle construction patterns, each having similar characteristic parameters that ran be easily extracted from the myoelectric signal. By using these parameters, it is possible to segregate different muscle contraction patterns into classes. Each class of muscle contraction is used to trigger a particular function in the prosthetic device. A neural network implementation is applied to myoelectric signal analysis tasks. The motivation behind this research is to explore more reliable methods of deriving control for a multidegree of freedom arm prosthesis. Autoregressive model parameters and signal power are used as features. >