Feature Selection And Application Based On Wlan Indoor Positioning System

Recently as many corporations like Google and Apple gradually pay attention to indoor positioning techniques, study on indoor positioning algorithms is lifted new upsurge. Location information is mainly used in Location Based Services especially based on mobile devices. LBS has been widely used in information delivering in big shopping mall, navigation in hospital, location navigation in airport, indoor positioning in large museums and entertainment venues. It is well known that with the development of GPS positioning method, outdoor positioning services has met the demands of people. However, location services based on indoor circumstance needs to be improved, especially in accuracy, reliability, real-time positioning ability and so on.Against this background, this paper firstly makes a generally discussion and comparison among indoor positioning systems at stage, such as GPS positioning system, Infrared positioning system, Radio Frequency Identification positioning system, Ultra-wideband positioning system, Bluetooth positioning system and WLAN positioning system and so on. This paper pays attentionto many aspects, especially in terms of positioning accuracy, reliability, real-time positioning ability and limitations. And then summarize the advantages and disadvantages of WLAN indoor positioning systems. After that, this paper gives some examples of the main three categories of positioning algorithms at present, highlight the superiority of fingerprinting methods, and introduce machine learning algorithms within fingerprinting indoor positioning area at the same time. Specially, an adaptive boosting algorithm, also called AdaBoost which is a combination of a series of weak classifiers like decision trees according to heuristic principles, is introduced in the traditional fingerprinting algorithms. AdaBoost is a classifier that not only has a fast real-time location capability, but also performs well with respect to location accuracy. Besides, feature selection algorithms also play an important role in indoor positioning in practice. According to the simulation in school building, feature selection could help to reduce model complexity, shorten model training cycle, while enhancing the prediction ability and fault tolerance of a model. In particular, with the combination of boosting algorithm and feature selection algorithms, indoor positioning accuracy will be enhanced, model training period will be shorten and real-time location ability will be improved.