A stable and adaptive pole-zero filter

Most adaptive problems of interest are non-linear, however, the application of non-linear, or pole-zero filter to these problems have been restricted due to the poor convergence properties of adaptive pole-zero algorithms that often result in an unstable solution. Here, we present a stable pole-zero filter model for such applications. In a given example, this filter resulted in significant, up to 4 times reduction in computations and up to 7 time reduction in memory bandwidth when compared to a linear filter for the same problem. The above algorithm has been used for echo cancellation and demonstrated in a directory assistance system that has been deployed in carrier networks and also for channel equalization in a V.17 modem(1).