Cepstrum Thresholding Scheme for Nonparametric Estimation of Smooth Spectra

Simple and efficient nonparametric spectral estimation is becoming more important as a signal processing tool in applied measurement science. A novel spectral analysis technique based on cepstrum thresholding is presented, which is shown to be an effective semi-automatic way of obtaining a smoothed nonparametric estimate of the spectrum of a stationary signal. The numerical study includes an example involving a moving average (MA) signal, where it is shown that the proposed spectral estimator outperforms significantly the standard periodogram