Deconvolution of marine seismic signal using higher order statistics and Bernoulli-Gaussian model

A solution to the deconvolution problem of high resolution underwater seismic reflection is proposed. The seismic source wavelet, is identified by a causal non-minimum phase ARMA model using higher order statistics (HOS) information, while the layered system, which is characterized by reflection coefficients and travel times, is assumed to be a Bernoulli-Gaussian process. The seismic trace is deconvolved by an iterative detection-estimation algorithm based on maximum likelihood approach. The whole method is applied to seismic traces coming from various underwater sources.