Research On Indoor Localization Algorithm Based On Particle Filter With Smartphone

With the maturity of IEEE802.11technology and the development of mobile Internet,Location Based Services has gradually penetrated into people’s daily lives, and plays avery important role on various occasions. It has broad application prospects in the medicalassistance, car-navigation and other fields. How to accurately obtain the locationinformation of the mobile user is an important element of Location-Based Services. Inorder to overcome the disadvantages of the existing indoor localization methods, and toprovide accurate and reliable location information to the mobile users in indoorenvironment, this paper proposes indoor localization method based on particle filter andfully draw on existing mature indoor localization technology. This method uses thepedestrian dead reckoning and ranging two kinds of wireless location technology formulti-mode integration, and it will be applied in the indoor environment pedestrian trackand localization. And this method builds its own fingerprint database also. We focus on thefollowing aspects.First, we propose pedestrian dead reckoning method based on phone built-in sensor.Through filtering and feature extraction the data collected by the axis acceleration sensorduring the pedestrian walking to detect pedestrian step. Smart phone’s in-builtlow-precision sensors can’t meet the accuracy requirements of reckoning. In order toavoid errors accumulate with time, we proposed pedestrian dead reckoning method basedon step detection.Secondly, we propose pedestrian localization method using particle filter. We studythe relationship between the received signal strength and the distance. And we find thepropagation model in indoor environment. For the essence of Dead Reckoning is vectorsum, the error will accumulate with time. In order to suppress the accumulation of error,this paper uses particle filter to fuse the localization data of Dead reckoning andpropagation model.Again, we build fingerprint database with localization result. Location and timeinformation is uploaded to the database. We incrementally build location fingerprint database. And this method eliminates offline training phase of the traditional fingerprintlocalization.Finally, we design experiments to prove the accuracy and feasibility of the proposedlocalization method. And the experimental results are analyzed.