A step-by-step modeling technique using a generalized adaptive IIR delay filter

A new modeling technique using adaptive IIR delay filters is presented. In each step of the adaptation process a significant delay parameter of the unknown system is estimated and a new parallel section which is comprised of one, two or four adaptive parameters is added to the previous sections. The number of the coefficients in each step is computed adaptively using a procedure that leads also to an improved initialization of these coefficients which are then adapted via the recursive Gauss-Newton algorithm. Computer simulation results are presented to illustrate the robust convergence properties of the proposed adaptive technique, particularly in modeling unknown systems with impulse responses that include sparse blocks, e.g., a multiple echo path. Because of the parallel and, at most, two-pole structure of the filter sections the stability monitoring requires essentially no computations and the gradient complexity in each adaptive step will be less than that of the parallel realization of adaptive IIR filters