Neuromorphic cross correlation of digital spreading codes

Abstract

The study of neural networks is inspired by the mystery of how the brain works. In a quest to solve this mystery, scientists and engineers hope that they will learn how to build more powerful computational systems that are capable of processing information much more efficiently than today’s digital computer systems. This dissertation involves a biologically inspired circuit which can be used as an alternative for a cross correlation engine. Cross correlation engines are widely used in spread spectrum, wireless communication systems that use digital spreading codes to divide a single communication medium into separate channels. This technology is used in many systems such as GPS, ZigBee and GSM mobile communications. The technology is renowned for its robustness and security since it is highly tolerant to signal jamming and spoofing. Digital spreading in wireless communication is also widely used in military systems and has recently been proposed for use in the medical sector for neural prostheses. A limitation of using digital spreading is that the computational demands on the cross correlation engine are normally quite high and is generally considered to be the limiting factor in designing low-power portable devices. In recent developments proposed by Tapson, it was shown that a two-neuron mutual inhibition network can be used to generate a cross correlation like function (Tapson et al., 2008). In this work, the two-neuron cross correlation engine is analysed specifically for application on a particular set of digital spreading codes called Gold codes. Based on the analysis, the neuron’s response to an input signal is optimised in favour of yielding a neural cross correlation that resembles the mathematical cross correlation more closely. The aim is to find a biologically inspired computer that is practically viable in an electrical engineering application involving a digital spread spectrum communication system.