A Recursive Autoregressive Method for Spectral Estimation

In this article, we present a parametric technique for estimating the frequency of sinusoids in noise. The method is autoregressive where the AR model parameters are found by solving the normal equations recursively. The proposed method differs from existing ones in that only few iterations (one or two) are used to estimate the model parameters. This method, which we herein refer to as RAMSE is not sensitive to the model order, does not exhibit spectral line splitting and does not generate spurious peaks.