Hits:

Indexed by:期刊论文

Date of Publication:2011-12-01

Journal:Journal of Information and Computational Science

Included Journals:EI、Scopus

Volume:8

Issue:14

Page Number:3035-3042

ISSN No.:15487741

Abstract:Correntropy is a localized similarity measure between two scalar random variables (RVs). In this paper, a new adaptive time delay estimation (TDE) algorithm is devised for the impulsive stable noise based on correntropy. It is known that in the TDE problem, the time difference between relevant sensors can be modeled as a finite impulsive response filter whose weight coefficients are samples of a sinc function. Unlike previously introduced least mean p-norm time delay estimation, the new algorithm obtained the weight coefficients of the filter by maximum correntropy criterion needing neither prior information nor estimation of the numerical value of the stable noise's characteristic exponent. In order to solve the kernel selection problem, we propose the procedure to find the kernel size adaptively by making use of the constraint that the sum of the squares of the weight coefficient of filter equals unity. Simulations show the proposed algorithm offers improved performance over the existing least mean square algorithm and least mean p-norm approach when the noise is highly impulsive at lower signal to noise ratio. 1548-7741/Copyright ? 2011 Binary Information Press December 2011.