Distributed Maximum Likelihood DOA Estimation Algorithm for Correlated Signals in Wireless Sensor Network

Abstract

The performance of the existing direction of arrival (DOA) estimation algorithms for source localization in the wireless sensor networks (WSN) degrades when the sources are correlated. The degradation is due to the rank deficiency of the source covariance matrix which is reflected in the high cramer-Rao lower bound (CRLB) for correlated sources. Unlike subspace based techniques, maximum likelihood (ML) based technique does not require any preprocessing technique for DOA estimation of correlated signals. Hence ML technique can be directly applied for WSN with arbitrary array geometry. The DOA estimation accuracy for correlated signals is improved by employing distributed ML approach in this paper. The subarray formed at a particular node which is experiencing highest CRLB can improve its estimation accuracy by receiving better DOA estimates from its neighbors. The corresponding distributed CRLB is derived and found a substantial improvement in the distributed scenario. The distributed CRLB lies between that of local CRLB for the subarray formed at the node and global CRLB for the array formed by all sensors in the WSN. Diffusion quantum particle swarm optimization is used to optimize ML estimator for fully correlated signals as it has only a single parameter for tuning. Simulation results show that the estimation accuracy improves at a node even for correlated signals yielding highest CRLB using distributed approach. The root mean square error at a specific node using distributed algorithm approaches to the derived distributed CRLB.