Convex Optimization Via Jensen-Bregman Divergence for WLAN Indoor Positioning System

In the last two decades, the indoor positioning systems (IPS) have attracted many researchers because of the great importance in many pervasive applications. Different techniques have been developed for IPS, a method that fingerprints the Received Signal Strength (RSS) of WLAN at specific places that can obtain high accuracy of about one meter at the exact location. A large range of indoor navigation needs can be provided by using IPS, especially in unusual conditions such as being in large complex buildings or emergency healthcare needs. IPS can play a great role in other applications that needs tracking and observing such as for the elderly people or for security purpose. In this paper, a framework that incorporates the probabilistic neural network (PNN) with Jensen-Bregman Divergence (JBD) is proposed. To validate our algorithm, the results were compared with PNN and kNN nearest neighbor. Where implemented inside an academic building. The experiment results show that PNN-JBD achieves competitive performance comparing with traditional approaches.