Bispectrum and reconstruction of modulation signal extracted from ship-radiated noise

I.IntroductionRecentyearshigh-ordcrstatisticshavebeguntofindwideapP1icabilityindiversefie1dssuchassonar,radar,plasmaphysics,biomedicine,seismicdataprocessing,harmonicretrieva1,adaptivefi1teringandarrayprocessingl'-'l=Thcsestatisticsnotonlyrevea1amp1itudeinforma-honaboutaprocess,butalsorevca1phaseinformation.Insigna1processing,thesearethreemotivationsbehindtheuseofhighcr-ordcrspectra:1.TosupprcssGaussiannoiseofunknownspectra1charactcristicsindctcction,paramcterestimation,andsignalreconstructi…

Bispectrum and reconstruction of modulation signal extracted from ship-radiated noise

The ship radiated noise is analyzed by higher-order spectra, which preserve the phase characteristics of signals. Phase information is suppressed in power spectrum and double fre quency power spectrum. Bispectrum analysis on ship noise's modulation signal reveals that the phase relationships among the harmonics of periodical envelope is regular. Because phase val ue's sharp change near the peaks stands for the harmonics in frequency domain, to find the ex act phase value of the peak is difficult. So interpolation in frequency domain is needed. After interpolation, there are more points in the narrow band at the peak, phase value's change be comes more smooth, and the peak's exact phase value can be obtained which meets our re quirement for it's precision. From the amplitude and the phase of these harmonics, thc period ical components of the envelope can be reconstructed, which can be called "Signature" of the ship noise. The signature can be used to recognize ships. Results of the analysis of real data show that this method is useful. One important discovery of our experiments is that the enve lope signals in all subbands of the ship-radiated noise are alike very much. This phenomenon can be utilized to devcloped a new approach of passive localization. It is a general knowledge that the sOund wave of different frequency propagates in shallow water at different group veloci ty. Ship noise's components in different subbands arrive the receiver at different time. The time delay can be obtained from the correlation function between the envelopes of the two com ponents. The time delay reflects the distance between the source and the receiver. One advan tage of this approach over the traditional is that only one sensor is needed instead of three.