Exact Formulation for Stochastic ML Estimation of DOA

Haihua CHEN  Masakiyo SUZUKI  

Publication
IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E93-A   No.11   pp.2141-2152
Publication Date: 2010/11/01
Online ISSN: 1745-1337
DOI: 10.1587/transfun.E93.A.2141
Print ISSN: 0916-8508
Type of Manuscript: Special Section PAPER (Special Section on Smart Multimedia & Communication Systems)
Category: 
Keyword: 
stochastic maximum likelihood (SML) estimation,  signal covariance matrix,  non-negative definiteness,  coherent signals,  global and local search methods,  

Full Text: PDF>>
Buy this Article




Summary: 
This paper addresses the issue of Unconditional or Stochastic Maximum likelihood (SML) estimation of directions-of-arrival (DOA) finding using sensors with arbitrary array configuration. The conventional SML estimation is formulated without an important condition that the covariance matrix of signal components must be non-negative definite. An likelihood function can not be evaluated exactly for all possible sets of directions. First, this paper reveals that the conventional SML has three problems due to the lack of the condition. 1) Solutions in the noise-free case are not unique. 2) Global solution in the noisy case becomes ambiguous occasionally. 3) There exist situations that any local solution does not satisfy the condition of the non-negative definiteness. We propose an exact formulation of the SML estimation of DOA to evaluate an likelihood function exactly for any possible set of directions. The proposed formulation can be utilized without any theoretical difficulty. The three problems of the conventional SML are solved by the proposed exact SML estimation. Furthermore we show a local search technique in the conventional SML has a good chance to find an optimal or suboptimal DOA although the suboptimal solutions violate the condition of the non-negative definiteness. Finally some simulation results are shown to demonstrate good estimation properties of the exact SML estimation.