ON FIR SYSTEM IDENTIFICATION FROM NOISY INPUT AND OUTPUT DATA

This paper is concerned with identifying parameters of finite impulse response (FIR) systems from noisy input-output data. The key idea is to estimate the input noise variance by minimizing a properly defined optimization criterion. Once a good estimate of the input noise variance is available, the unbiased estimates of the FIR system parameters are readily obtained by a closed-form least-squares solution without involving any iteration process. The proposed modified least-squares algorithm is compared with other existing methods through computer simulations.