2019•2019 1st International Conference on Advances in Science, Engineering and Robotics Technology (ICASERT)Requires access
Identification of Autoregressive Systems in the Presence of Additive Noise Using the Matrix Pencil Method
Md. Shamim Hussain
Open publisher page 0 citations
Abstract
In this paper, a novel method for estimation of the parameters of autoregressive (AR) systems in the presence of additive noise is presented. This method models the autocorrelation function (ACF) of the noisy AR signal as a sum of non-increasing complex exponentials and then estimates its parameters using the matrix pencil method. We improve upon this basic method by fitting the noisy AR signal to a higher order AR model and using its parameters to calculate a parametric ACF of the noisy signal which makes the pole extraction process using the matrix pencil method stable and more accurate. We show that this method can accurately estimate the parameters of an AR system in noise in a single pass even at a very low (-5dB) signal-to-noise ratio (SNR).