In this paper, we consider the fast direction-of-arrival (DOA) estimation f
or a spatially dispersed source. Using a less detailed model for the part o
f the covariance matrix that depends on the angular spread, we show that DO
A estimation can be decoupled from angular spread estimation. This results
in a one-dimensional minimization problem which can be solved using the fas
t Fourier transform. The so-obtained DOA estimate does not depend on any as
sumption on the spatial distribution of the source and is, hence, robust to
mismodeling. Both theoretical analysis and numerical simulations are prese
nted to illustrate the performance of the method. (C)1999 Academic Press.