S k C I This paper treats a signal detection problem using arctic under-ice noise. The authors have had access to one large segment of data (6150144 samples), which is nonstationary and has been shown to be non-Gaussian. A model is presented for the arctic under-ice noise, and the performance levels achieved by several different detectors are compared. The association between the shape of the empirical probability density function and the shape of the power spectrum is explored. The arctic noise is known to contain narrowband and impulsive components, and it is shown here that removal of the narrowband components results in nearly Gaussian noise.

Motivated b y the recurring use of the generalized Gaussian family to model different underwater noise sources a n d t h e asymptotic performance levels of some commonly used detectors for this family, we examined the performance of these detectors for several different underwater noise sources. T h e sources considered are non-Gaussian, highly correlated, generally nonstationary, a n d vary from being lighter to heavier tailed t h a n Gaussian. Although t h e linear detector had t h e best performance, the linear rank detector a n d the L-level uniform quantizer consistently h a d similar performance levels. Considering t h e simplicity a n d t h e robustness of the quantizer a n d the r a n k detectors, either of these detectors would seem to be the best choice for these environments.

No abstract

The aim of the work is to determine and clarify the effects of some typical problems that arise when working with a real-noise source. To this end, stationarity, time correlation between data samples, and parametric and nonparametric detection for a large segment of Arctic under-ice noise are examined. In particular, the effect that correlation between samples has on the ability to detect a known signal in additive noise using several different parametric detectors is analyzed. Since most underwater environments are both nonstationary and non-Gaussian, nonparametric detectors are of interest also. Thus a comparison of the performance levels of several different nonparametric detectors is performed for four common underwater environments.