A probability neural network-Jensen-Shannon divergence for a fingerprint based localization

Abstract

For decades, humans have been keen on creating smart spaces where advanced technology is utilized to provide enhanced services. Receiving directions and/or being recognized within indoor spaces is one feature of smart spaces that is currently heavily researched. Indoor positioning systems (IPS) can be used to provide a wide range of user navigation and directions services, particularly in abnormal conditions such as needing emergency healthcare services and being in unfamiliar complex buildings where some may become disoriented or lost. IPS also can be a friendly tool for people with vision impairment to allow for better livable communities for them. Other applications for IPS fall under tracking applications which may include activity recognition for security purposes and observation for the elderly or infirmed individuals. An indoor positioning system can be a hybrid system that uses multiple technologies such as wireless LAN, vision via cameras, motion sensors, or lasers to name few. In this paper we propose a technique for IPS using WiFi. The technique is based on a probabilistic neural network (PNN) scheme in which we incorporate the Jensen-Shannon divergence method. To validate our proposed method, we compare our results with the nearest neighbor method. Results indicate that our integrated system outperforms this method in terms of nearest neighbor estimation. Our results show that this method has the ability to achieve less than 1m accuracy in an academic building.