The Research Of IIR Adaptive Filtering

Nowadays the subject of adaptive filtering constitutes an important part of the statistical signal processing. It has two kinds of forms. One is the Finite Impulse Response filter, the other is the Infinite Impulse Response filter. Comparing to the former, the later has some advantages: It can offer the better performance than the Finite Impulse Response filter which has the same coefficients. That is to say, the coefficients that Infinite Impulse Response filter demands are less than the Finite Impulse Recursive filter in order to get the same performance. So because of the reduce of the computation and the progress of the performance, IIR filter becomes more and more popular in the field of Signal Processing. In the article, we focus on the some unbiased adaptive IIR filtering algorithm.Systems optimized under the Gaussian assumption often yield unacceptable performance when subjected to impulsive non-Gaussian noise. So in this dissertation, a class of distributions called alpha-stable distributions can be used to model these types of impulse noise, we give two kinds of adaptive algorithms, and we apply them to the IIR adaptive filtering.Estimation of the impulse of an unknown system from its measured input and output has found many applications in signal processing, communications and control. Many estimation algorithm based on least squares and least mean squares techniques have been developed for accurate system parameter estimation if additive noise exists only at the output.However, noise-free input measurements are difficult to obtain in many practical situations because these signals are also corrupted by quantization noise, self-noise, and/or other additive noise, and neglecting the presence of input noise will lead to biased impulse response estimation.At the last of the article, we give a new LMS algorithm based on the classical LMS algorithm. Because of its fast convergence and reduction of computation, it can be used in the field of adaptive IIR filtering.