Gradient Vectors Orthogonalization Based Adaptive Filtering Algorithm for Correlation Function
Gao Ying①②; Xie Sheng-li②
①Dept. of Computer Sci. and Tech., Guangzhou Univ.,Guangzhou 510405 China;②Dept. of Electron and Comm. Eng.,South China Univ. of Tech., Guangzhou 510641 China
Abstract：In this paper, the time varying forgetting factor exponentially weighted least square error criterion Jr(n)=∑nt=0 λn-teT(n)Ce(n),is proposed by reducing the correlation function mean square error criterion Jr(n) = E[eT(n)Ce(n)] proposed by Asharif(1999). Applying gradient method on the proposed criterion, a new adaptive filtering algorithm for correlation function is proposed by orthogonalizing the current gradient vector to the previous gradient vector. Computer simulation results indicate that the proposed algorithm’s convergence performance is better than that of ECLMS algorithm proposed by Asharif.
2004, Vol. 26 
