Comparison of quadratic phase coupling detectors on sonar data

Detection and estimation of processes with mixed spectra and coupled harmonics is a well-known problem in the analysis of time-series and is of particular interest in classification of sonar signals. In this paper a comparison is made between two quadratic phase coupling (QPC) detection techniques analytically. One technique adopts a segment and average approach to obtain estimates of the signal bicoherence and then uses a step-wise outlier rejection method to select only relevant data points. The other technique is a single record phase coupling detector which removes the need for multiple independent records. It is shown that analytically the techniques are identical.