A correlation domain algorithm for autoregressive system identification from noisy observations

This paper presents an identification technique for minimum-phase autoregressive (AR) systems using noise-corrupted observations. In order to reduce the effect of noise in the correlation domain, instead of using the conventional autocorrelation function (ACF), a once-repeated ACF (ORACF) of noisy observations has been employed. Based on characteristics of the ORACF under a noisy condition, a set of equations has been developed. The AR parameters are estimated by solving these equations in the form of a quadratic eigenvalue problem. Computer simulations are carried out for AR systems of different orders under noisy environments showing a superior identification performance in terms of estimation accuracy and consistency.