A new iterative method for solving total least squares problem

In this paper, we propose a new method to solve TLS problem adaptively, which uses the estimate of the generalized Rayleigh quotient, for adaptive FIR filtering. An approximation method to estimate the generalized Rayleigh quotient and its recursive version are presented and compared. The computer simulation results show that the proposed algorithm outperforms the RLS algorithm in adaptive system identification with noisy input circumstances and the recursive structure is superior to the direct approximation in the sense of stability.