Spectral Correlation Density Estimation Using 2N-FFT Interpolation

Many artificial signals exhibit second-order cyclostationarity which could be characterized in the frequency domain by means of the spectral correlation function or spectral correlation density (SCD). This paper introduces the approach to designing the time-smoothing SCD estimator that processes a finite-length discrete signal to evaluate SCD values at the centers of nodes of the rectangular grid on the bifrequency plane. The structure of the support region on the plane is discussed, and the difference in the width of the resolution cells along the conventional frequency and cyclic frequency axes is pointed out. The enough density of the grid to cover the bispectral plane completely is achieved by using FFT with the zero-padding interpolation technique, where the base is extended to the double length of the sequence. It prevents the loss of any cyclic components of the signal under processing. The simulation results show the performance of the proposed estimator applied to the examples of well-known signals possessing strong wide-sense cyclostationarity.