Direction of Arrival Estimation by Employing Intra-block Correlations in Sparse Bayesian Learning Through Covariance Model

Estimation of the arriving signal directions at the receiver side is of utmost important in the field areas of array signal processing. The proposed technique in this paper involves two major steps, in that the first step is a C-step where, we deduce the covariance model for Direction of Arrival (DOA) estimation and through which, the noise variance of the model will be estimated. In the second step, i.e. L-step, the covariance model deduced in the C-step will be used along with the noise statistics to estimate the variance of sparse DOA spectrum, which is unknown. In this step, Sparse Bayesian Learning with Expectation maximization framework is extended to exploit the property of intra-block correlations in the unknown DOA spectrum. The variance of sparse DOA spectrum, which is estimated in L-step indicates the locations of non-zero values in the spectrum, hence resulting in directions of the signal sources. In the results section, it can be seen that the increase in accuracy and performance of the proposed algorithm is one of the result of exploiting intra-block correlations. The covariance modelling in C-step results in high probability of true DOA estimation in the case where number of signal sources is less than the antenna elements in the Uniform Linear Array (ULA) with lesser number of snapshots required. It is also shown in the simulation results that an acceptable estimation accuracy is achieved in the case where number of signal sources is greater than or equal to the antenna elements, but with larger snapshots required.