Gibbs Sampling based Sparse Bayesian Learning for Direction-of-Arrival Estimation with Impulse Noise Towards 6G

Integrated communication and localization presents a highly promising prospect in the sixth-generation (6G) mobile communication systems. As its core element, direction-of-arrival (DOA) estimation is confronted with a challenging environment characterized by impulse noise. Conventional DOA estimation techniques fail to fully account for the negative impact of this detrimental noise, leading to a significant decline in estimation accuracy. A promising approach to solving this problem is sparse Bayesian learning (SBL). However, the application of SBL with large-scale antenna arrays is hindered by the high computational complexity. To address this challenge, an optimized Gibbs sampling based sparse Bayesian learning (GSSBL) algorithm is proposed by introducing an inverse-free high-dimensional sampler, by which the computational complexity with largescale antennas can be alleviated. Simulation results demonstrate that the proposed GSSBL approach can achieve accurate DOA estimation with lower complexity comparing with other state-of-the-art methods.