Artiical Neural Nets for Peripheral Nervous System -remoted Limb Prostheses

In this paper we review the aim of the INTER 1-project (Intelligent Neural InTERface), especially from the point of view of Artiicial Neural Nets (ANN). We also propose a modus operandi to process real nerve signals using ANN. Finally we present rst experiences using this proposal. Microfabricated neural interfaces promise to become a powerful tool for applications like the control of motor/sensory limb prostheses for amputees and the direct stimulation of spinal cord injuries (FNS). Such interfaces have been successfully implanted and signals of the peripheral nerve system (PNS) have been recorded. These signals have to be processed for remoting limb prostheses. We propose to interprete recorded signals using the Independent Components Analysis (INCA) to separate the incoming signals. We calculate the information dimension of the data to determine the optimal topology for self-organising map and use a self-organising map to classify the incoming signals. In a further state, signals from sensors on the prostheses will be processed using ANNs in order to transmit information to the PNS.