Distributed signal detection in underwater multi-array systems with partial spatial coherence

In recent years, using multiple arrays instead of a single array has raised great interest in various applications. In this paper, we consider a distributed signal detection problem using an underwater multi-array system. Beamforming is first performed at each hydrophone array to increase local array output signal-to-noise ratio (SNR). Then these arrays are viewed as 'virtual' sensor nodes in an underwater acoustic sensor network, and beamformer outputs are considered as the received signal samples at these nodes. We also assume imperfect spatial coherence across the sensor nodes, which is modeled as a multiplicative noise process. The decentralized implementation is then achieved by computing the largest eigenvalue of the sample covariance at the individual arrays using a decentralized implementation of the power method. Then we apply the well-known Roy's largest root test and check the significance of the largest eigenvalue, which acts as a constant false alarm rate (CFAR) detector. With recent results in random matrix theory (RMT), the threshold of the proposed detector is decided by the distribution of the largest eigenvalue of a pure noise matrix which depends on the number of sensor nodes and samples. Numerical simulations were done to demonstrate the effectiveness of the proposed signal detector. Comparisons were also made on the performance of the proposed detector with perfect and partial spatial coherence.