A technique for improving detection and estimation of signals contaminated by under ice noise

Recent analyses of FRAM II arctic data have shown that under ice ambient noise can be at times highly impulsive and non-Gaussian. The analyses included time domain statistical measurements which were consistent with previously reported results of experiments made within the Canadian Arctic Archipelago. New findings in the frequency domain based on skew, kurtosis, and cumulative distribution function estimates, also indicate the existence of strong non-Gaussian noise. It is known that the ability to detect and estimate signals contaminated with non-Gaussian noise using conventional processing is degraded compared with optimum techniques which utilize knowledge of the noise statistics. Results comparing the performance of conventional and nearly optimum signal processing methods are presented using the FRAM II data.