Study on synthesis of low coefficient sensitivity digtial filter structures using genetic programming

摘要

This paper presents a method to synthesize digital filter structures of second-order transfer functions using Genetic Programming (GP). Filter structures are represented by S-expressions. Applying genetic operators to the S-expressions generates various new S-expressions. Each S-expression is given a fitness value based on the coefficient sensitivity. Thus, GP can synthesize low coefficient sensitivity digital filters. A numerical example shows that our proposed method can synthesize filter structures with very low Sensitivity compared with the low coefficient sensitivity state-space realization.