On the Consistency of Likelihood Penalization Methods in Large Sensor Networks
The problem of detection, i.e., estimating the number of sources from noisy observations is a fundamental problem in signal processing, as well as in many other fields like geophysics, finance or biomedical engineering. In the context of array processing, detection is performed by using N observations collected by an array of M sensors. In general, detection is a first step to obtain more precise informations on source localization (Direction of Arrival estimation), power of noise and source signals, or to perform advanced techniques like source separation. Numerous methods have been proposed since the past 40 years, and in particular, Likelihood Penalization (LP) methods, which are among the most popular.