Comparative study for optimisation of causal IIR perfect reconstruction filter banks

Comparative study for optimisation of causal infinite impulse response (IIR) filters with applications to perfect reconstruction quadrature mirror filter (QMF) banks is presented. A hybrid procedure is applied where a constrained genetic algorithm is first used to search the objective function landscape for a promising valley. Sub-optimal filter parameters are then further optimised using four different methods. These are: new GA based `creep code´, gradient based constrained Sequential Quadratic programming method and Quasi-Newton method and non-gradient based downhill Simplex method. Finally, a comparison is drawn between each of the four methods of optimisation