Our research investigates a filter-bank framework combined with a sparse-subspace method for Direction- of-Arrival (DoA) estimation in 5G indoor localization. We propose Filter-Bank Signal Subspace Orthogonal Matching Pursuit (FB-SS-OMP), a method tailored to DFT-s-OFDM, the uplink technique employed in multi-user 5G networks. Using ray-tracing simulations in a virtual indoor environment and composed of multiple transmitters. We empirically show that conventional DoA estimators, 2D-MUSIC and 2D-ESPRIT, degrade severely in coherent multipath conditions typical of indoor channels. In contrast, their filter-bank counterparts maintain substantial high accuracy and can resolve more DoAs than the spatial degree of freedom of such conventional methods. FB-SS-OMP achieves the best performance, improving angular accuracy by approximately 0.3° to 1.1° compared with filter-bank 2D-MUSIC and filter-bank 2D-ESPRIT.