Liu Xiushan

Weak signal detection and asymptotic relative efficiency in Gaussian-Exponential mixture noise

Traditional signal detectors for underwater targets are based on Gaussian model. Such model, according to recent research, does not agree with actual ocean ambient noise, and the optimal detector based on this model risks unexpectedly poor performance. We present a Gaussian- Exponential mixture model to describe the received noise by analysis of underwater noise samples for deriving an optimal signal detector. We first rebuild the ocean ambient noise model as a Gaussian-Exponential mixture model. Next, we derive an optimal nonlinear detector structure for known signals in such noise. Finally, in order to evaluate the performance of the optimal detector, it is compared to the conventional energy detector. The comparison is based on the concept of asymptotic relative efficiency (ARE) for comparing hypothesis testing procedures. The simulation results show that the performance of the proposed detector is better than that of the conventional energy detector by 4∼8dB.