Decimation-based frequency estimation using nonlinear prefiltering in frequency domain

In this paper, the problem of estimating the frequencies of closely spaced complex exponentials in the presence of colored noise is considered. In order to get rid of the effect of colored noise from the observations before we use ESPRIT method to estimate the frequencies, we first use the nonlinear prefiltering in frequency domain, which is based on the observation that the exponentials in the colored noise show themselves as outliers in frequency domain. While the ESPRIT method performs poorly when applied to closely spaced exponentials, decimation is proposed as a way to increase the performance in the case of over sampling. Simulation results show the efficiency of the nonlinear prefiltering in frequency domain and the estimation accuracy by using the decimation technique.