A Model of Silicon Neurons Suitable For Speech Recognition

Abstract

In this paper we describe the structure of an artificial neuron designed using only basic
electronic components such as transistors, resistors, capacitors and diodes which could be successfully
implemented as the computation unit for a neural analogue integrated chip. Being able to associate
between temporal events and to extract similarities from physical data, these neurons could be used as a
tool for speech recognition or image processing problems. To test the performance of the electrical neural
network under real conditions it was developed a tool which could perform speaker independent vowel
recognition. The final goal of this approach is to manufacture an analogue integrated chip with similar
behaviour as the neocortical column which is the basic functional unit of the biological brain.