Design of optimal IIR filters in reduced-dimensionality parameter spaces

A new method of designing WLS optimal IIR filters is proposed in this paper. The approach belongs to the emerging group of methods that search for the optimal filter parameters inside the spaces of significantly reduced dimensionalities. Efficient implementation of the proposed method is achieved by selecting suitable parameterisation of the filter's transfer function and deriving numerical formulas for calculating the WLS cost and its derivatives with respect to the optimisation parameters. The filters can thus be designed using off-the-shelf programmes tackling nonlinear optimisation problems with constraints. Owing to the reduction of the dimensionality of the search space the proposed approach is more robust and provides more accurate results than the “fully blown” traditional techniques of designing IIR filters.