Performance analysis of cascade spline adaptive filtering based on normalized orthogonal gradient adaptive algorithm

Abstrak

In this paper, the cascade architecture of spline adaptive filtering (CSAF) for nonlinear systems is presented with the normalized version of orthogonal gradient adaptive (NOGA) algorithm. Spline adaptive filtering comprises a sandwich of the first linear adaptive filtering (LAF) and nonlinear adaptive look-up table. In this cascading architecture, SAF is connected to the second LAF. NOGA is considered as the fast convergence applied by stochastic gradient-based approach. Convergence properties of the proposed NOGA-CSAF algorithm in terms of instantaneous errors can be derived by using Taylor series expansion. Experimental results demonstrate the effectiveness of proposed NOGA-CSAF algorithm using the mean square error scheme. It clearly outperforms the traditional least mean square algorithm on CSAF model in the nonlinear identification system.