Improved parameter estimation of linear systems with noisy data

This paper addresses the problem of parameter estimation of linear systems with noisy input-output measurements. A new and simple estimation scheme for the variances of the white input and output measurement noises is presented which is based on expanding the denominator polynomial of the system transfer function only and makes no use of the average least-squares (LS) errors. The attractive feature of the iterative LS based parametric algorithm thus developed is its improved convergence property. The effectiveness of the developed identification algorithm is demonstrated through numerical illustrations.