Sample Fourth-order Cumulant Tensor Denoising for DOA Estimation with Coprime L-shaped Array

Tensor-based direction-of-arrival (DOA) estimation using sparse arrays suffers performance loss from noise perturbations which are twofold, namely, the noise power embedded in the auto-correlation-based covariance statistics, and the noise introduced by sample statistical calculation. In this paper, a fourth-order cumulant tensor-based DOA estimation method for a specially designed coprime L-shaped array is proposed, where both noise perturbations are effectively addressed. The proposed method utilizes the cross-correlation statistics to avoid generating the noise power, based on which a fourth-order cumulant tensor is formulated for coarray signal processing. Then, to attenuate the sample fourth-order noise, its statistical property is investigated for an adaptive noise attenuation threshold design. Finally, a dual directional concatenation approach is developed to obtain a denoised fourth-order structural coarray tensor for underdetermined DOA estimation. The effectiveness of the proposed method is verified by simulations.