Signal analysis using spectral correlation measurement

Abstract

The spectral correlation analysis of cyclostationary signals can be considered as a substantial extension of the well known time averaged periodogram spectral analysis since it detects periodically with time varying second order moments. The most general description of cyclostationary signals is based on the Wigner-Ville spectrum (WVS). The periodic structure of the WVS results in discrete slices of its Fourier transform, the bifrequent spectral correlation function that indicates spectral coherence. An efficient estimation procedure for the spectral correlation based on a FFT of a sequence of pseudo Wigner distributions is given including a search efficient procedure for discrete cycle frequency detection. An implementation based on a multi-DSP platform is described as well. The spectral correlation analysis can be preferably applied for signals that are produced by some periodic modification or modulation of stationary random noise. Therefore, many applications for signal analysis in technical fields can be found, especially for modulated signals in communications as well as for noise and vibration signals produced by rotating machines.