Application of nonarithmetic filtering techniques for improved system identification

System identification and parameter estimation may be significantly impacted by the presence of noise. The amount and type of noise present in a system are important factors, and perhaps the most significant impact of noise is on the accuracy and robustness of system models. Noise reduction by pre-filtering signals provides system identification schemes with "cleaner" data from which more accurate system models may be constructed. However, traditional filtering methods may also introduce noise. Recently, a class of nonarithmetic filters has been developed that has proven successful in eliminating impulsive and Gaussian distributed noise. These nonarithmetic filters are applied to system identification strategies to reduce the effects of signal contamination on parameter estimation.