DOA estimation in the presence of unknown colored noise, the global matched filter approach

Abstract

The problem of the localization of multiple narrow band sources in the presence of arbitrary noise of unknown spatial spectral density is addressed. The array geometry can be arbitrary but must be known. The spatial noise spectrum is described using a sufficiently rich class of models that somehow covers the set of rational spectra. The Global Matched Filter is used to identify the characteristics of the sources that are present and to get a approximate model of the unknown colored noise. It is a technique that can be seen as a model-fitting or sparse representation approach in which the observations are decomposed on the association of different bases of candidate models. The computational complexity is reasonable and the performance are quite good and compare favorably with other methods.