DOA Estimation Aided by Magnitude Measurements

Abstract

This correspondence discusses the estimation of direction-of-arrival (DOA) in a magnitude-aided antenna array (MA-AA), where magnitude-only radio frequency (RF) chains are introduced into the classical AA to acquire magnitude measurements. DOAs are initially estimated by the multiple signal classification (MUSIC) algorithm based on the complex-valued measurements from the conventional antennas. After griding the neighborhoods of these initial DOAs, the DOA estimation problem with the hybrid observations in MA-AA is converted as the recovery problem of sparse signals, which can be resolved by generalized approximate message passing (GAMP). Due to the “spatial leakage” effect, non-zero clusters appear around true DOAs. Their positions (i.e., non-zero supports) provide DOAs estimations. Moreover, DOAs and supports remain unchanged in several snapshots, then common supports are shared. Therefore, the cluster-sparse property of sparse signals is exploited by modeling a hidden Markov-tree (HMT) in the shared supports, on which belief propagation (BP) is executed to recover the binary probabilities of supports. Some unknown hyper-parameters in GAMP and BP are learned by expectation-maximization (EM). In comparison to existing estimators, EM-BP-GAMP shows advantages on DOA estimation, computational complexity, and DOA resolution. With the EM-BP-GAMP estimator, MA-AA is more energy-efficient than the classical AA. These advantages are successfully validated by experimental results.