Research On Direction Of Arrival Estimation Algorithms For Wideband Signals Based On Variable Inference

Direction of arrival(DOA)estimation is widely applied in positioning,especially in indoor positioning scenarios,which can availably avoid the impact of multipath and improve the precision and robustness of indoor positioning.The traditional DOA estimation algorithms have certain limitations in practical application,because of the worse performance at small snapshots and low signal-to-noise ratio.Sparse Bayesian learning(SBL)algorithms based on sparse representation improve the accuracy of DOA estimation and reduce the computational complexity.In addition,most DOA estimation methods are aimed at narrowband signals,which are not suitable for wideband signals.The selection of a single frequency will seriously affect the estimation performance.The sparse Bayesian learning(SBL)methods based on sparse representation have higher accuracy under lower complexity,and provide a new aspect for DOA estimation.This paper focuses on the wideband DOA estimation algorithm within the SBL framework.The main research work and innovation points of this paper are as follows:A novel on-grid DOA estimation method is proposed for wideband signals with Dirichlet process(DP)prior,which adopts the uncommon assumption that the signals from different orientations occupy the different spectral bands.The DP prior is introduced to cluster the sub-bands with different sparsity patterns.A factor graph is designed to describe the relationship between variables.According to the distribution of the variables,we employ the combined belief propagation-mean field(BP-MF)message passing rule on the factor graph to infer the hidden variables,and lead to the novel algorithm for wideband DOA estimation.Through simulation experiments,we found that the algorithm is more precise on-grid DOA estimations than other state-of-art methods.The on-grid algorithms inevitably have quantization errors,due to the limited discrete division of the space domain.In order to deal with the problem of grid mismatch,this paper proposes a combined BP-MF algorithm for wideband off-grid DOA estimation.In this paper,the off-grid angle deviation parameter is introduced to describe the distance between the true DOA and the nearest grid angle,and an off-grid DOA estimation model is established.According to the relationship between variables,this paper redesigns the factor graph,and applies the combined BP-MF rule and expectation maximization(EM)algorithm to inference the latent variables and parameter.Simulation results show that the off-grid algorithm for broadband signals proposed in this paper can effectively reduce the errors caused by grid mismatch,and the estimation accuracy is closest to the Cramer-Rao lower bounds(CRLB).In addition,the algorithm proposed in this paper has better performance by comparing with other algorithms under the conditions of different number of sources,grid division,selected sub-bands and snapshots.