A neural network characterisation of electromyography. Part I

In this paper we seek to examine the extent to which motor unit firing sequences can be related to EMG signal sequences leading to the characterisation of tasks from their associated EMG temporal patterns. Arising from this, the work described here is concerned with the development of a computer model to simulate the muscular activity during elbow flexion. To achieve this we start by expanding upon models of the central nervous system (CNS) and the process of trajectory formation coupled with the use of a Neural Network to predict muscular activation, based on prerecorded EMG activation patterns for known tasks. In this way we can generate a pattern of muscular activation from physiological and kinematic data required to perform a task sequence. The EMG data, used in the networks training, has been taken from males aged between 18 to 26 years all without any known disabilities or injuries to the elbow joint. This paper reports on data processing and event detection used in this experiment and suggests further work in this field.