Efficient Design of Cosine-Modulated Filter Banks Using Evolutionary Multi-objective Optimization

We propose a novel and efficient way to design maximally decimated FIR cosine modulated filter banks, in which each analysis and synthesis filter has linear phase. We consider a class of near-perfect reconstruction CMFBs with the linear phase prototype filter, which structurally eliminates the amplitude overall distortion. The prototype filter design problem is then formulated into a multi-objective optimization problem (MOP), which aims at maximizing stop-band attenuation and minimizing reconstruction error simultaneously. We have modeled the design problem as a constrained multi-objective optimization problem which is efficiently solved by using a recently proposed algorithm MOEA/DFD. Experiment shows that the performance of MOEA/DFD exceeds that of MOEA/D and NSGA-II.