Summary
International Symposium on Nonlinear Theory and Its Applications
2015
Session Number:C3L-A
Session:
Number:C3L-A-3
Design Framework for Practical Optimum Nonlinear Filter for Unknown Noise Characteristics
Yukihiro Tadokoro,
pp.891-894
Publication Date:2015/12/1
Online ISSN:2188-5079
Summary:
A nonlinear filtering method was previously proposed that can estimate a weak signal buried in strong non-Gaussian noise. This method is useful in signal processing because it can realize the Cramer-Rao lower bound in the estimation. To determine the filter characteristics, a mathematical expression for the probability density function (PDF) of the noise is necessary. The original study assumed that the PDF was known, but in practical situations, this assumption is not true. The present study considers the use of the filter in a situation where the PDF is unknown. A method for estimating the PDF is presented, together with the corresponding filter function. Kernel density estimation is employed using Epanechnikov kernel in order to reduce the computational complexity, and the optimum bandwidth (at which the estimation performance is maximized) is derived. Through numerical evaluation, the proposed method is confirmed to be effective.