Canonical signed digit representation of Quadrature Mirror Filter using Genetic Algorithm

In this paper a two channel Quadrature Mirror Filter (QMF) bank is designed using Genetic Algorithm (GA) and coefficients of filters are represented in Canonical Signed Digit (CSD). A fresh Genetic operator is used that combines crossover and adaptive mutation to improve the convergence rate and solution quality of GA. Prototype filter is constructed using GA and the coefficients of the filter are converted to nearest signed power of two terms which reduces hardware complexity of the overall system. The proposed design of QMF is also achieved least value of Reconstruction error. The technique improves the performance of QMF with respect to its peak reconstruction error, transition bandwidth and hardware required to realize FIR digital filter to less than the complexity of 2's complement representation. During the optimization of prototype filter coefficients only nonzero CSD digits are used so that better performance is achieved.