Wideband signal detection and maximum-likelihood source localization

We present a signal detection scheme and maximum-likelihood source localization for wideband signals using a randomly distributed sensor array. The signal detection is based on a simple statistical test that sequentially detects the number of sources, or model order, with a theoretical probability of false alarm. The detection scheme requires the maximum-likelihood parameter estimation, in which an efficient alternating projection procedure is also introduced, to detect a model order greater than two. A model order correction step is introduced to reduce overestimation of model order due to the error in the maximum-likelihood estimates. The proposed scheme is evaluated via computer simulations and experimental data.