Using generalized similarity filter to enhance proximity detection for sparse beacon deployment

Considering an incomplete signals acquisition due to a sparse beacon deployment, this paper proposes a generalized similarity filter to improve the performance of proximity detection and thus guarantee the quality of proximity-based service (PBS). In particular, this paper leverages Bluetooth Low Energy (BLE) Beacons to realize a PBS system which comprises a number of Proximities of Interest (PoIs). We define a PoI as an object or area which is associated with a beacon such that each PoI can announce their presence implicitly through the beacon's signal. However, under a sparse beacon network condition in which some beacons associated with some PoIs are malfunction or their batteries die before the scheduled maintenance, a receiver (e.g., smartphone) might fail to return the target PoI correctly. In view of the quality degradation in consequence to the sparse condition, we refine the performance of classical compressive sensing based algorithm with a generalized similarity filter. The effects of different similarity measures on proximity detection performance are also investigated. Simulation results indicate that the proposed algorithm improves the detection accuracy as compared to the conventional compressive sensing based algorithm. Specifically, Chordal-based similarity filter achieves substantial improvement in comparison with Mahalanobis and Euclidean-based similarity computation.