| This thesis presents an enhanced version of a Genetic Algorithm to design discrete filter coefficients. This optimization based algorithm has the advantages of eliminating truncation error and quantization process in filter design, and to produce filters which are suitable for high speed signal processing applications. We have examined the usefulness of various error norms, such as Least Mean Square and Minimax, and their impact on the convergence rate and the result. We also present the application of various encoding schemes applied to the filter coefficients and their effect on obtaining optimized filters. Examples of 1-D & 2-D FIR and 2-D IIR filters are provided to illustrate the design procedures and to determine the best Genetic Algorithm combinations for digital filter design. |
