New Adaptive IIR Filtering Algorithms.
ILLINOIS UNIV AT URBANA COORDINATED SCIENCE LAB
Pagination or Media Count:
A family of adaptive IIR filtering algorithms is proposed based on the Steiglitz-McBride identification scheme. The algorithms are shown to be close approximations of one another for slow adaptation. Because of the non-vanishing gain, they are suitable for filtering applications and are simple to implement. A convergence proof is carried out using a theorem of wide-sense convergence in probability in the literature of stochastic processes. For the sufficient order case, the estimates can be shown to converge to the true values. While for the case of reduced order, it is conjectured that the estimates converge to the best fit, which is supported by computer simulations. The major drawback is that the estimates may be biased in presence of colored disturbance. However, this does not restrict the applicability of the proposed algorithms to some important practical problems. One specific topic, adaptive echo canceling, is extensively studied and simulated for various situations. The results are favorable compared with the conventional adaptive FIR cancelers and other adaptive IIR algorithms. Author
- Statistics and Probability