New frequency and 2-D DOA estimation algorithm

This paper proposes a new algorithm to jointly estimate the frequencies and two-dimensional (2-D) directions-of-arrival (DOA) of multiple narrowband sources. Firstly, this algorithm constructs four cumulant matrices using the fourth-order cumulants of some properly chosen sensor outputs, and secondly, unlike the conventional parallel factor (PARAFAC) model in the data or subspace domain, it forms a three-way array (TWA) in the cumulant domain, and analyzes the uniqueness of the low-rank TWA decomposition. Thirdly, it jointly estimates the signal parameters from four diagonal matrices that are obtained from the low-rank decomposition of this TWA. The proposed method requires neither pairing parameters nor search computation, and can be applied to arbitrary Gaussian noise environment. Simulation results are presented to validate the performance of the proposed method.