Enhancing Wi-Fi based indoor positioning using fingerprinting methods by implementing neural networks algorithm in real environment

Abstract

Global positioning systems have difficulties in finding positions inside buildings, since indoor positioning needs additional indoor infrastructures deployment. In this research, indoor positioning by using Wi-Fi access point is investigated as the main usage of Location Based Service (LBS) applications. We employed fingerprinting method to increase the accuracy of positioning. The study has been done in real environment in Universiti Teknologi Malaysia (UTM). Two models were designed by using Neural Network algorithm for indoor positioning. The fingerprinting dataset contains received signal strength from different numbers of existing Wi-Fi access points in the real environment. Accuracy rate and mean square error were calculated for the algorithm. Evaluations of models have been done by conducting experiments to compare both models. Analysis suggests that Neural Network method which achieved 71% of accuracy with number of neurons = 11 is the most precise model for indoor positioning in this project. In future, more features can be applied to this model in order to increase the accuracy. This approach has the potential to be implemented as a real mobile application for indoor environment.