Adaptive Filters and Neural Net

Abstract

A filter is a physical hardware or software system that is used to extract useful information from noisy input data. Linear and nonlinear filters are two classes of filters. The adjective “adaptive” refers to a system that tries to adjust its parameters to meet specific objectives which depend on the state of the filter system as well as the state of system environment. It is achieved using a recursive/iterative algorithm, which helps the filter to operate efficiently in an environment where there is no absolute information about the required quantities. In this chapter, we briefly discuss about various kinds of adaptive filters and different applications of the same in the areas of communication systems, biomedical signal processing, noise/echo cancellation systems, and Radar applications. The performance of various adaptive filtering algorithms can be compared in terms of elapsed time and mean squared error (MSE) vs. number of iterations. The MSE of LMS and RLS filters vary randomly, but MSE of NLMS filter decreases gradually. An artificial neural network contains interconnection of a huge number of nonlinear electronic units called neurons. The development of neural networks has been motivated by the structure and functioning of human brain, hence its name. In this chapter, we briefly discuss about single- and multi-layer neural net and applications of the same in the areas of telecommunication systems, biomedical applications, and signal processing.