Detection of Gaussian signals on a sensor array

In array signal processing, the most common model describing the received signals and noise is Gaussian. It is well known that if the model is proper, the estimator which gives the best performance is the one constructed by fully using the statistic information of the model. The authors develop a detector for the number of signals. Such an estimator is constructed with utilising the log-likelihood function and an information theoretical criterion. The consistency of the detector is proved and the simulations with a comparison to the MDL detector are also presented.<>