In this paper, an extended UN-ESPRIT (Wu and Reilly, 1991) method is p
roposed to estimate the direction of arrival (DOA) in unknown noise fi
elds. Compared with the UN-ESPRIT approach, the new method can be used
to perform high resolution DOA estimation with simpler structure of a
rrays and less sensors by using virtual interpolated arrays. On the as
sumption that the noise correlation is spatially limited, the effects
of unknown noise are eliminated through canonical correlation analysis
of the data received from two spatially separated real arrays, and DO
A estimation is carried out from the outputs of two interpolated array
s, which are translationally equivalent and have arbitrary array geome
tries.