Improving indoor positioning system by using wi-fi fingerprint with machine learning methods

As devices (including smartphones, Personal Digital assistants, netbooks etc) are been equipped with state of the art technology making them susceptible to new generation Wi-Fi networks, getting more integrated and getting us more interconnected. Because of this, most mobile devices has brought with it the possibility of providing a lot of various services that utilize a cheap network infrastructure. Existing approaches such as Bluetooth and GPS signal services are not sufficient enough to meet the requirements of accuracy and flexibility. In contrast, Wi-Fi, which is commonly available on most smartphones, provides a compelling alternative for proximity estimation using machine learning tools. In this thesis, we demonstrate through experimental studies the efficacy of Wi-Fi for this exact purpose. We propose a proximity estimation model to determine the distance based on the RSS values of Wi-Fi data in 3 different buildings, each having at up to 4 floors. Thesis has explored a Wi-Fi fingerprint based localization scheme that exploits the power of machine learning methods to solving the problems of Wi-Fi fingerprint localization. The data used was acquired from UJIndoorLoc dataset wish is publicly available and the recognition rate are impressive.