Learning-Enhanced Joint Estimation of AOAs and Source Number With Quantized Phase-Only Measurements

This correspondence proposes to estimate the angle of arrival (AOA) and source number with quantized phase-only (PO) measurements extracted via multiple one-bit analog-to-digital converters (ADCs), thereby significantly reducing the power consumption. A density-based spatial clustering of applications with noise (DBSCAN-) enhanced expectation-maximization (EM-) generalized approximate message passing (GAMP-) based estimator is developed. Firstly, the AOA estimation problem is converted as detecting supports of cluster-sparse signals and then solved by modified EM-GAMP in a single snapshot. Secondly, the coarse AOA estimates from multiple snapshots are clustered by DBSCAN to estimate the source number and improve the AOA estimation accuracy. Simulation results show that the quantized PO measurements scheme is more energy-efficient than the conventional complex-valued measurements scheme. The AOA and source number estimation performance of this scheme is superior to that of the one-bit quantized measurements scheme due to the extreme quantization loss of the latter. Furthermore, the DBSCAN-enhanced estimator incorporates the AOA estimate results from multiple snapshots and effectively eliminates outlier AOA estimates, thereby improving the AOA estimation performance, particularly at low signal-to-noise ratios.