Identification of noisy AR systems using damped sinusoidal model of autocorrelation function

This letter presents a novel method for minimum-phase autoregressive (AR) system identification at a very low SNR using damped sinusoidal model representation of the autocorrelation function of the noise-free AR signal with guaranteed stability. The new model parameters are estimated solely from the given noisy observations. Then AR parameters are obtained directly from the estimates of the damped sinusoidal model parameters. The simulation results show that the proposed method can estimate the AR system parameters with high accuracy even at an SNR as low as -5dB.