Comparative analysis of digital filter optimization using parallel genetic algorithm

Digital filters are used to modify some characteristics of digital signals by means of a series of multiplication and addition operations. Optimization of the filter is realized to determine the optimal filter coefficients that provide the desired characteristics. Filter design with traditional searching techniques to achieve coefficients may remain trapped in a local minima point. In Genetic algorithm, global minimum can be found via continuing search operation from different points and thus optimal values can be determined. In Genetic algorithms, when the size of the problem to be solved increases, parallel realization which is one of the most effective ways applied for the acceleration of the algorithm. In this study, performance analysis of digital filters optimization using parallel genetic algorithms was carried out on multi-core computer. For this purpose, parallel genetic algorithm was coded with C# programming language and built in Parallel library was used for parallel computations. Performance for different filter structures was examined experimentally and the duration of the sequential optimization algorithm is reduced by parallel algorithm. When the experimental measurements with four and six cores processors using developed interface are compared with sequential implementation it is observed that performance is increased depending on the number of cores.