Online Kernel Canonical Correlation Analysis for Supervised Equalization of Wiener Systems

We consider the application of kernel canonical correlation analysis (K-CCA) to the supervised equalization of Wiener systems. Although a considerable amount of research has been carried out on identification/equalization of Wiener models, in this paper we show that K-CCA is a particularly suitable technique for the inversion of these nonlinear dynamic systems. Another contribution of this paper is the development of an online K-CCA algorithm which combines a sliding-window approach with a recently proposed reformulation of CCA as an iterative regression problem. This online algorithm permits fast equalization of time-varying Wiener systems. Simulation examples are added to illustrate the performance of the proposed method.