Fast communication
Wideband DOA estimation from the sparse recovery perspective
for the spatial-only modeling of array data
Nan Hu, Dongyang Xu, Xu Xu, Zhongfu Ye
n
National Engineering Laboratory for Speech and Language Information Processing and Institute of Statistical Signal Processing, Department of Electronic
Engineering and Information Science, University of Science and Technology of China,
Hefei, Anhui 230027, China
article info
Article history:
Received 28 June 2011
Received in revised form
28 November 2011
Accepted 2 December 2011
Available online 9 December 2011
Keywords:
Direction-of-arrival estimation
Wideband source
Least absolute shrinkage and selection
operator
Second-order cone programming
abstract
This communication utilizes a sparse recovery technique named ‘‘least absol ute
shrinkage and selection operator’’ (LASSO) to formulate and solve the problem of
direction-of-arrival (DOA) estimation for far-field wideband sources. The spatial-only
modeling for wideband array output is employed when the sources have flat spectra,
and the DOA estimation problem in this model can be transformed into a sparse
recovery problem. Via the LASSO te chnique, an optimization problem is constructed
and solved by second-order con e (SOC) programming to obtain the DOA estimates.
Numerical simulations are also provided in contrast with those existing algorithms
using the spatial-only model.
& 2011 Elsevier B.V. All rights reserved.
1. Introduction
As a problem which arises in various applications such
as radar, sonar, radio-astronomy, etc., direction-of-arrival
(DOA) estimation for far-field sources has been an active
research topic in the past decades [1]. Most of the
algorithms proposed for this problem are generally tai-
lored for the narrowband case, and for wideband cases the
conventional way is to split the array outputs into several
narrowband signals by passing them through a bank of
narrowband filters and then combining the source loca-
tion information embedded in these bands [2–4].
Incoherent signal subspace method (ISSM) [2] is a
straightforward way to tackle the multi-band information
aggregation problem, which incoherently combines the
results of independently performed estimations at multiple
frequencies. ISSM suffers from the unreliable DOA
estimates in some low signal-to-noise ratio (SNR) frequency
bands and the incoherent style of information aggregation
prevents it from resolving coherent sources. Coherent
signal subspace method (CSSM) [3,4] replaces a set of
estimation problems with a single one through focusing
the covariance matrices of all the bands to that of a certain
band. CSSM is capable of resolving coherent sources due to
frequency smoothing in the procedure of averaging the
subspace-aligned covariance matrices, and substantially
improves the resolution capability and estimation accuracy
in low SNR. However, the fatal weakness of CSSM is that the
performance of this method relies on the accuracy of
preliminary DOA estimates or focusing range of the specific
angular sector. The methods mentioned above can all be
summarized as spatial–temporal approaches, which need
the frequency decomposition using narrowband filters.
Given the fact that the output of a narrowband filter which
has a long impulse response is highly correlated in time, we
have to acquire a long data record to give good estimates
through spatial–temporal approaches. The spatial-only
model [5,6] provides an alternative way of wideband DOA
Contents lists available at SciVerse ScienceDirect
journal homepage: www.elsevier.com/locate/sigpro
Signal Proce ssing
0165-1684/$ - see front matter & 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.sigpro.2011.12.002
n
Corresponding author. Tel.: þ86 0551 3601314.
E-mail address: yezf@ustc.edu.cn (Z. Ye).
Signal Processing 92 (2012) 1359–1364