ABSTRACT
Adaptive filters have wide applications in data communication, system identification, spectrum analysis, adaptive beamforming, magnetic recording, image pro-cessing etc. Adaptive filters learn the characteristics of a signal as it is processed and tend to approach the performance of an optimal filter for the given application[1]. The well known adaptation algorithms include the least-mean-square (LMS) [2] algorithm and the recursive least-squares (RLS) algorithm [3],[4]. Traditionally, LMS is the more commonly used algorithm in adaptive filters. The LMS algorithm is an approximation of the steepest descent method [5]. Instead of estimating the cross-correlation and auto-correlation matrices from the data, the instantaneous values of the quantities are used. The LMS algorithm converges to an optimum filter over a period of time. The resulting algorithm is very simple to implement and robust. The LMS algorithm is well understood by the community in the industry. Efficient structures are available for implementation of the LMS algorithm. Blind equalization techniques using LMS algorithm are also well developed [6],[7]. Joint equalization and carrier recovery schemes have been implemented succesfully [8]. These advancements make LMS suitable for practical applications.
