Broadband matched field processing using robust prewhitening and multiple window techniques

The estimation of source and environmental parameters for stationary broadband wave propagation in an unknown shallow ocean is considered. We apply an approximate maximum-likelihood technique (AMLE) in the frequency domain after cleaning the observations from outliers using robust prewhitening. This method is compared with the previously used multitaper techniques. Environmental noise in data obtained by horizontally towed arrays just beneath the ocean surface has been observed to be more spiky than Gaussian due to the sound originating from individual breaking waves. Therefore, we apply the robust-resistant spectral methods introduced by Martin and Thomson (1982). The advantage of this approach is the physically and statistically justified full broadband estimator. Simulations show the feasibility of the presented approach. Finally, the algorithm is applied to real sonar data recorded by a horizontally towed receiver array in the Baltic Sea in order to estimate the ocean depth and source bearings and to interpret the impinging signals