Angle estimation based on Vandermonde constrained CP tensor decomposition for bistatic MIMO radar under spatially colored noise.

• The problem of Vandermonde constrained CP tensor decomposition for angle estimation in MIMO radar under spatially colored noise is addressed. • The smoothed estimate of the temporal cross-correlation matrix is presented to provide improved robustness to colored noise in the case of the limited number of pulses. • Constrained alternating least squares (ALS) based on new rectification strategy for Vandermonde factors is used to solve CP decomposition model. • With the help of temporally smoothed cross-correlation and constrained ALS, the proposed method exhibits superior angle estimation performance. We address the problem of Vandermonde constrained CANDECOMP/PARAFAC (CP) tensor decomposition in application to angle estimation for bistatic multiple-input multiple output (MIMO) radar under spatially colored noise. By exploiting the temporally uncorrelated characteristic of colored noise, a new denoising scheme based on the temporally smoothed cross-correlation approach is presented. Then, after rearranging the smoothed cross-correlation matrix into a fourth-order tensor, a Vandermonde constrained CP tensor decomposition model is formulated, which fully exploits the multidimensional structure of the array measurement and the Vandermonde structure of the factor matrices. To solve this model, an efficient constrained alternating least squares (ALS) algorithm is developed to decompose the Vandermonde factor matrices. Finally, joint estimates of direction of departure (DOD) and direction of arrival (DOA) are obtained by the generators of the Vandermonde factor matrices. Compared with the state-of-the-art approaches, the proposed method can achieve better angle estimation performance by jointly using the temporally smoothed cross-correlation denoising operation and enforcing the Vandermonde structure information in the tensor decomposition. Numerical simulation results verify the effectiveness and improvement of our algorithm. [ABSTRACT FROM AUTHOR]