Power spectral Density Estimation via Overlapping Nonlinear Averaging

This paper proposes a new method for accurate estimation of power spectral density (psd) of noisy signals. The major drawback caused by the tradeoff between spectral resolution and variance is examined. The main aim of this paper is to show how the accuracy of the periodogram can be improved by adopting nonlinear averaging techniques of partially overlapped sampled data. In addition, the effects of the taper function and the percentage overlap are investigated with respect to the computational cost and the achievable improvements. The derived expressions for the bias and variance indicate the better performance of this method with respect to the previous ones. Finally, experimental validation of such results is shown.