Modified RLS algorithm for unbiased estimation of FIR system with input and output noise

In the presence of both input and output noise, the classical least squares solution for finite-impulse response (FIR) estimation is biased. It has been shown that bias can be removed by properly scaling the optimal FIR filter coefficients in the least-squares (LS) criterion. A modified recursive least squares (MRLS) algorithm is proposed for accurate identification of a system with both input and output noise. Simulation results show that this method outperforms the modified LMS algorithm under non-stationary interference conditions