A Research On 2-D DOA Estimation Algorithm Based On Sparse Bayesian Learning In Massive MIMO System

The application of massive MIMO(Multiple-Input Multiple-Output)technology in 5G mobile communications has led to significant performance improvements in mobile communication systems,especially in terms of transmission rate and capacity gain,which is closely related to the Direction of arrival(DOA).DOA is an important research area in the direction of array signals and is often applied in radar,biomedical,sonar,navigation,tracking of various objects and rescue,etc.Most of the traditional DOA estimation algorithms are based on subspace class algorithms,which require high signal-to-noise ratio and large number of snapshots to obtain good estimates,when the signal is coherent,the estimation results are seriously affected.In recent years,with the development of sparse theory,the emergence of sparse reconstruction algorithm provides a new idea for DOA estimation research.Sparse algorithm can recover the original signal accurately from the data sampled at a much lower sampling rate than Nyquist,which addressed the problems of low estimation accuracy and slow convergence of traditional algorithms,and that makes the research of DOA estimation algorithm based on sparse theory very meaningful in massive MIMO system.Massive MIMO systems deploy huge antenna arrays at base stations,and the array signal processing faces complex computational problems.To better compensate for the parameter estimation deficiencies of traditional algorithms in massive MIMO systems,sparse reconstruction algorithms are used widely,which mainly include convex optimization algorithms,greedy algorithms,and sparse Bayesian algorithms,the principle of algorithm is that sparse the received signal data and reconstruct the target signal with a small amount of data.The convex optimization algorithm requires a computational optimization tool,resulting in the rapid increase of system calculation.While the greedy algorithm does not guarantee the global optimum of the iterative process and has weak noise immunity,which makes the estimation accuracy is not global optimal solution.Comparing the above,the sparse Bayesian learning algorithm not only can guarantee the global optimal,but has better noise immunity under the conditions of small number of snapshot and low signal-to-noise ratio.In summary,in order to improve the timeliness and reliability of DOA estimation in massive MIMO systems,we propose an improved sparse Bayesian learning algorithm,in which the signal reception matrix is processed by singular value decomposition,so as to speed up the convergence of the algorithm.Secondly,the offgrid model is used to reduce the modeling error of the traditional SBL(Sparse Bayesian learning)algorithm,and updated the grid parameters through the first-order Taylor approximation to improve the estimation performance of the algorithm,but the proposed algorithm does not completely eliminate modeling error.Therefore,we propose a polynomial-based DOA estimation algorithm,and treat the coarse grid points as adjustable parameters and updating the grid parameters by calculating the roots of polynomial,so that to eliminate the modeling error.In addition,due to subspace leakage,most conventional algorithms will lead to the loss of sampling data in harsh and complex massive MIMO environment,we introduce pseudo-noise resampling method to enhance accurate of algorithms in the complex massive MIMO environment.The performance of the improved algorithm is verified by the experimental simulation analysis of the algorithm.