New structures and algorithms for adaptive system identification and channel equalization

Abstract

The main drawback of the ADF is that it takes lot of iteration and fails to identify nonlinear systems. BAF converges fast while maintaining the same performance as ADF but its performance degrades at nonlinear conditions.In this thesis we propose an ANN, which provides better and faster converges when employed for identifying nonlinear systems. This network employs chebyschev based nonlinear inputs updated with the RLS algorithm. Through extensive computer simulation it is demonstrated that CFLANN updated with RLS is a better candidate compared to FLANN and MLP in terms of less complex structure, less number of input simple needed and does accurate identification.