Bispectral analysis of underwater EM pulses

In this paper we investigate the higher-order spectral properties of an electromagnetic (EM) transmission channel. We use higher-order moments ensemble averaging of received real underwater EM pulses for signal to noise enhancement and discuss appropriate features based on bicoherence and the cross-bispectrum for object classification. The reconstructed signal y/spl tilde/(n) is estimated from the bicepstrum. This technique is shown to outperform conventional stacking techniques. We use bicoherence for estimation of nonlinear properties in the received pulses. There are no significant quadratic nonlinear contributions in the signal. The non-linearities are essential around the power supply frequency and at a region above 17 kHz. Finally an analysis of the transfer function based on the cross-bispectrum is given. At several frequencies in the signal bandwidth there is a significant part with nonlinear gain.