Research On Key Indoor Positioning Technologies Based On Mobile Terminal

With the thrive and development of location service, localization technique, as a precondition of location service, has drawn more and more attention. Global Positioning System (GPS), as a widely used localization technique, can provide quite accurate outdoor positioning result. However, GPS does not work in indoor scenario for its signal cannot penetrate building well.With the ubiquitous of smart mobile terminal, indoor localization based on mobile terminal faces a broad application prospect. Among current researches in the field of indoor positioning, localization technique based on fingerprint and dead reckoning technique are the two most common technologies applied in different situations:(1) Wireless signal distribution in localization area is known. Fingerprint localization, based on received signal strength (RSS), receives much attention for its low cost and easy deployment. Fingerprint localization needs to build up a fingerprint database called radio map before positioning, and define a location as positioning result by matching current wireless signal to records in fingerprint database. Due to this procedure, fingerprint localization cannot be applied in area where signal distribution is unknown. To assure accurate positioning result, numerous data sampling and calibration work is needed to build up the radio map.(2) Wireless signal distribution in localization area is unknown. Dead reckoning can position mobile terminal by utilizing the inertial information, which is collected by the inertial sensor embedded in mobile terminal. Positioning error will accumulate with inertial sensor measure error. Thus, positioning accuracy will decline dramatically along with time.To improve accuracy and efficiency of localization, this paper studies different enhancements for current key indoor localization techniques based on mobile terminal applied in the two above-mentioned situations. According to wireless signal’s space variation pattern, this paper applies Gaussian process regression in fingerprint localization for efficient radio map building method by reducing sampling density. This paper applies particle filter technique combining with fingerprint localization in dead reckoning method, to solve the problem that inertial positioning error accumulates with observation measurement error.In this paper, the corresponding validation experiments for the above location technologies are conducted and the results show that:Gaussian process regression model can well predict the wireless signal received strength of spatial location. Gaussian process regression model can be used to predict signal strength across the positioning area for the establishment of fingerprint database, reducing the signal sample collection density; Particle filter technology combining location fingerprint can suppress error cumulative effect of inertial navigation.