Design of high throughput recursive and non-recursive digital filters in one and two dimensions with Canonic Signed Digit coefficients and sub-expression elimination using Genetic Algorithm

In this dissertation methods of obtaining high throughput rate digital filters are examined. The use of Canonic Signed Digit (CSD) filter coefficients is established and a new chromosome coding technique is developed to enable efficient design of non-recursive filters using a Genetic Algorithm.; The new genetic algorithm approach using the proposed new coding scheme is extended to efficiently handle recursive filters using a new unstable penalty factor to handle the instability constraints imposed by such filters. A technique is presented that allows these new methods to be applied to the design of high throughput 2-D filters.; The throughput rate of CSD coefficients digital filters is further increased by the use of common sub-expression elimination. A new graphical transformation is presented that allows for optimization of the elimination of CSD-coefficient common sub-expressions in both the vertical and horizontal dimensions.; The effectiveness of the proposed methods is demonstrated with example designs and comparisons to other methods.