Kernel-Based APA Adaptive Filters for Complex Data
Kernel-based adaptive filters present a new opportunity to re-cast nonlinear optimization problems over a RKHS and transform the original nonlinear task into a linear one, where one may employ linear and well-known adaptive algorithms. It also allows different types of nonlinearities to be treated in a unifying way. This method has been used in the machine learning community for some time. Now, its use in adaptive filters elevates the subject of adaptive filtering theory to a new level, presenting a new designing methodology of nonlinear adaptive filters. We have recently applied the complex Gaussian kernel to develop Affine Projection (AP)-based complex algorithms for Kernel Adaptive Filtering. In this paper, we now consider the performance of these algorithms for different input signal distributions, specifically the circularity of the complex input. Recent work has shown the use of widely-linear filtering leads to use of complete second-order statistical data when performing mean square estimation for complex data. We also extend recently developed complex kernel-based algorithms using the idea of widely-linear (WL) estimation. Simulations are used to verify our theoretical results.
Tokunbo Ogunfunmi Thomas Paul
Santa Clara University, Santa Clara, CA 95053, USA
国际会议
2011亚太信号与信息处理协会年度峰会(APSIPAASC 2011)
西安
英文
1-9
2011-10-18（万方平台首次上网日期，不代表论文的发表时间）