Signal-to-noise Ratio Enhancement Of Cyclic Summation

Abstract

Many signals encountered in passive sonar are approximately cyclo-stationary in that a given waveform is repeated at almost a constant rate. The underlying physical process causing the sound is often a complex train of nonidentical impulses. The ability to track the frequency of these pulses over time can be used in such things as detection, classification and target motion analysis. Methods based on high resolution Fourier analysis generally assume that the signal can be represented by a coherent sum over a large number of harmonic components, any one of which can be tracked over time. However there can be advantages in tracking at a time resolution sufficient to detect the pulses in time domain. One is that many harmonics can be tracked in parallel rather than individually. Another is better system identification by detecting the different underlying pulse trains. The contribution here is to evaluate the implications of cyclic summations for the signal-to-noise ratio, and demonstrate its application to frequency tracking and system identification.