Abstract: To tackle the poor accuracy issue in far-filed wideband signal direction of arrival (DOA) estimation under low signal-to-noise ratio conditions, a wideband DOA estimation algorithm is proposed based on tensor domain denoising. First, a tensor is constructed with the data from each sub-band, and then, high-order singular value decomposition is performed on this tensor to separate signals and noise by the minimum description length criterion. Next, the covariance matrix fitting algorithm is improved, and the signal power is constrained by using L1-norm to obtain solving model. Finally, the wideband signal DOA can be obtained by fusing all the narrowband estimation results. Simulation results demonstrate that the proposed algorithm can effectively reduce noise. Compared with root multiple signal classification algorithm and estimation of signal parameters via rotational invariance technique algorithm, the proposed algorithm does not neet to know the number of signals in advance for DOA estimation, and has low root mean square error under low signal-to-noise ratio conditions.

Key words: direction of arrival, low signal-to-noise ratio, high-order singular value decomposition, covariance matrix fitting