Optimization and adaptation of discrete-valued digital filter parameters by simulated annealing

A novel method is presented for the optimal choice of discrete-valued digital filter parameters (e.g. digital filter length). The method proposed is a “running” version of the simulated annealing algorithm. Its convergence properties are studied. An application of the proposed method to the adaptation of the median filter length is presented. Simulation results prove the convergence of proposed technique. It is also proven that simulating annealing algorithms give better results than iterative improvement (hill climbing) algorithms