WiFi Fingerprinting for Indoor Room Localization Based on CRF Prediction

Reports have shown that people spend 70% of their time indoor on average. Understanding indoor activities and locations have become significant research issues in recent years. Due to the massive deployment of WiFi access points around the world, the feasibility of indoor localization based on WiFi signals has been justified. There has been a lot of research work providing acceptable accuracy for room localization, which is either too complicated at building WiFi fingerprints and acquires the assistance of additional information or needs a considerable number of WiFi scans to provide localization. In this paper, we propose the Conditional Random Field Predication based Room Fingerprinting (CRF-PF) to predict the fingerprints of rooms that are not yet visited by referencing established room fingerprints. Experimental results show that compare to existing approaches, CRF-PF is more accurate on room localization by 12%.