| The use of an adaptive filter offers an attractive solution to the problem of processing signals that result from any operation of unknown statistics. Adaptive filtering is used for various applications such as linear prediction, channel equalization, adaptive noise and echo cancellation, and adaptive modeling and system identification. A wide variety of algorithms for the operation of adaptive filters have been developed in the past few decades. The choice of an algorithm is determined by various factors such as rate of convergence, robustness, computational requirements, numerical properties, and algorithmic structure. During the past decade, multiplierless filters have received widespread attention in the literature because of their high computational speed and low cost implementation. A multiplierless filter uses power-of-2 coefficients, which require only shifters and adders for filtering the data. Several design methods for multiplierless filters have been also developed. However, the work done on multiplierless adaptive filters has been very minimal.; This research focuses on developing multiplierless algorithms for the adaptive filtering for one-dimensional and two-dimensional signal processing. The potential use of the well-known genetic algorithm is explored in the design of an adaptive multiplierless filter. The research also focuses on the development of a quantized LMS algorithm using power-of-2 coefficients. A new way of multidimensional filtering for hyperspectral images is also introduced in the thesis. In the thesis, we also propose new ways to enhance images using edge detection schemes. |
