A novel unitary PARAFAC method for DOD and DOA estimation in bistatic MIMO radar

• Published in: Signal processing Vol. 138; pp. 273 - 279
• Main Authors: Xu, Baoqing, Zhao, Yongbo, Cheng, Zengfei, Li, Hui
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2017
• Subjects: Angle estimation; Multiple-input multiple-output radar; Parallel factor decomposition; Real-valued tensor; Real-valued tensor; Angle estimation; Multiple-input multiple-output radar; Parallel factor decomposition
• ISSN: 0165-1684; 1872-7557
• DOI: 10.1016/j.sigpro.2017.03.016

•The proposed unitary parallel factor (U-PARAFAC) algorithm is based on tensor decomposition.•The real-valued tensor signal model still follows a PARAFAC model.•Traditional unitary ESPRIT method is firstly extended to the real-valued PARAFAC model.•The proposed U-PARAFAC algorithm directly operates the real-valued loading factors instead of estimating the signal subspace. In this paper, a novel unitary parallel factor (U-PARAFAC) algorithm of estimating direction-of-departure (DOD) and direction-of-arrival (DOA) in bistatic multiple-input multiple-output (MIMO) radar is proposed. A real-valued tensor signal model is constructed by applying the traditional forward-backward averaging technique. Subsequently, the fact that the real-valued tensor follows a PARAFAC model is proved, thus the subspace-based high-order singular value decomposition (HOSVD) method can be avoided in the subsequent solving process. Furthermore, directly operating the real-valued loading factors instead of the signal subspace, traditional unitary ESPRIT (U-ESPRIT) method is firstly extended to the real-valued PARAFAC model. The new algorithm, which exploits the multidimensional structure and does not require the estimation of signal subspace, having good performance especially at low signal-to-noise ratio (SNR). More attractively, compared with classical tensor methods such as the PARAFAC algorithm and the unitary tensor-ESPRIT algorithm, the U-PARAFAC algorithm still performs well without sacrificing array aperture when targets are highly correlated or closely spaced. Additional angle pair-matching is not required. Simulation results verify the effectiveness of the proposed algorithm.