Akademik Veri Yönetim Sistemi
Klasik ve Modern Yön Kestirimi, Temel Engin Tuncer,Benjamin Friedlander, Editör, Elsevier Science, Oxford/Amsterdam , London, ss.1-15, 2009
Nonuniform linear arrays (NLAs) have certain advantages and problems for
directions of arrival (DOA) estimation. They cover a large array
aperture with fewer sensors, and they require fewer matched channels or
receivers. Their disadvantage is that coherent sources cannot be easily
handled. Furthermore, they need additional computation to compensate for
and augment the missing sensor information. The completion of missing
sensor data is required to improve accuracy. Array mapping is an
effective method of augmenting the NLA covariance matrix. Array-mapping
accuracy can be improved significantly if an initial DOA estimate is
used and then the estimates are improved iteratively. For this reason,
initial DOA estimation is a key problem for NLA, but it can be easily
solved for uncorrelated sources. Toeplitz completion can be used
directly for this purpose. Initial DOA estimation for coherent sources
is not an easy task. A promising approach for coherent signals is to use
partly filled NLA. Initial DOA estimates can be obtained via
forward-backward spatial smoothing for the ULA part of this array. Then,
array mapping can generate a covariance matrix corresponding to a full
array with the same aperture. Different array-mapping techniques exist
in the literature. Classical array interpolation is well known, but it
has certain limitations. Wiener array interpolation performs well, but
it produces focusing loss for NLA. It performs better for circular
arrays compared to alternative techniques where a large angular sector
for array mapping is used.