Blind equalization using neural networks and higher order statistics

Abstract

Blind equalization has been one of the most active areas of research in recent years. The potential application of blind equalization in wireless communication is one of the main reasons for its popularity. This thesis compares four different methods of blind equalization for nonminimum phase systems. Two Higher Order Statistics algorithms are used for channel identification. The first one is the Optimization al-gorithm and the second is Overdetermined Recursive Instrumental Variable (ORIV) algorithm. Two kinds of neural networks are used as equalizers to recover the trans-mitted signal. One is Multilayer Feedforward Network (MFN) based on Backpropa-gation algorithm, the other is Minimal Resource Allocation Network (MRAN) which is a newly developed Radial Basis Function Network that produces a parsimonious network structure.