An input-balanced realization based adaptive IIR filter using particle swarm optimization

In this paper, based on input-balanced realizations (IBR) and the particle swarm optimization (PSO) technique a novel adaptive IIR filter is proposed. This filter is derived from the input-balanced realization (IBR) that yields an excellent performance against finite precision errors. With such a realization, the stability of the adaptive filter can be ensured easily. As well known, the traditional gradient based adaptive algorithms cannot jump out of local minima of a cost function. To overcome this problem, the PSO, one of the intelligent optimization techniques, is employed. As the filter is implemented in a state space realization, we formulate a new cost function to include the PSO in the adaptive filter. This process is universal on any state-space realization based filters. Numerical examples show that the proposed adaptive filter yields a satisfactory performance.