An efficient approach to unbiased identification of noisy FIR systems

The problem of parameter estimation of finite impulse response (FIR) systems from noisy input-output data is investigated. An auxiliary matrix is introduced to formulate an eigen problem. The input noise variance is then estimated by solving the constructed eigen problem for a particular eigenvalue. With this, unbiased estimates of the FIR coefficients as well as the output noise variance can be readily obtained. The developed approach can be implemented at a low numerical cost. Computer simulations are given to demonstrate the good performance of the developed approach in comparison with other existing methods.