EM based GLRT Detector for MIMO Active Sonar using Bistatic Reverberation Model

It is known that the performance of sonar is limited by reverberation, especially, in shallow waters. We study the active MIMO (Multiple Input Multiple Output) sonar detection problem in scenarios limited by reverberation. The bistatic reverberation envelop at the hydrophone of the MIMO array is modeled as delayed and attenuated returns of the transmitted signals from the scattering patches at the ocean floor. Based on this bistatic reverberation ocean channel model, we derive a single snapshot generalized likelihood ratio test (GLRT) detector for MIMO sonar. In the proposed detection algorithm, an EM (Expectation-Maximization) based technique is employed for estimating the scattering strength coefficients in the reverberation model. The theoretical limits on the achievable performance of GLRT and the detection threshold are derived for the narrowband case. The performance of the proposed algorithm is compared with other techniques such as covariance matching and matched filtering. Simulations show that the performance of the proposed detector is quite close to the theoretical limit and outperforms other detection techniques.