Detection of short signal in a noisy underwater environment

Detecting underwater activity is a vast and important challenge. Whether it is for biological or economical reasons it is important to be able to correctly distinguish the many kinds of entities which belong to animal domain or artificial object domain. Most of the time, noisy environment coupled with a modelization problem make underwater detection a difficult problem. In order to avoid a fastidious work consisting on to provide a consistent model of signal of interest, we propose in this study to use the stochastic matched filtering theory in a detection context. As a matter of fact, this processing allows to detect useful information in a noisy environment even with unfavorable signal to noise ratio, under the assumptions that second order statistics of signal of interest and noise are known. Among all the things that should be interesting to detect, we focus on records of clicks of whales in a noisy environment. We show, using this filtering approach in detection, it is possible to detect clicks even with -8 dB unfavorable signal to noise ratio. A comparison with the classical matched filter is done.