Research On KNN-PIT Indoor Adaptive Fingerprint Location Technology Based On Wi-Fi

With the rapid development of mobile communication and Internet technology, the application need of location-based services are becoming increasingly strong. Since GNSS can provide continuous, high-precision of outdoor location information, to achieve lots of outdoor location services, such as vehicle, pedestrian navigation and positioning. But in the complex and volatile environment, GNSS can't navigate or locate because of signal weakening or failure. So the high-precision of indoor positioning technology has become a hot research topic.Therefore, the indoor positioning scheme using short range wireless communication has emerged. Because the Wi-Fi is a wireless communication standard, has long transmission distance, high fidelity, strong mobility, easy networking features, and has been widely deployed in large indoor public places, based on Wi-Fi fingerprint positioning technology has become the first choice in indoor LBS applications.But the current Wi-Fi fingerprint positioning scheme still have some problems to be resolved, Such as the signal in indoor propagation by effect of multipath and NLOS has time-varying characteristic, that will affect the reliability of the positioning; The reference point density affects the complexity of the algorithm and affect the real-time positioning; the traditional KNN algorithm can only roughly estimate the position of the target point range, cannot further constraints to target point range. In views of these problems, an improved KNN-PIT indoor positioning algorithm is put forward. The main work and innovations are as follows:(1)According to structure feature of indoor space,establishing the location fingerprint database which has class labels. The traditional fingerprint database only contains the location and the corresponding received signal strength indicator vector. Adding class label for location fingerprint and wireless access point can reduce the matching area and reduce the computational complexity.(2)Introduce virtual reference points,use theory of PIT to further constrain localization area of target point,use positioning algorithm adaptively to carry out positioning. Virtual reference points are not in the real fingerprint database, they are assumed when positioning with the KNN algorithm. The use of virtual reference points not only help to improve the positioning accuracy, also help to reduce the capacity of fingerprint database, reduce the computational complexity.(3)Comprehensively use Gauss filtering and mean filtering to reduce random errors of signal in online and offline stages.In the off-line phase, Gauss filtering is used to deal with the large sample of Wi-Fi signal, and remove large interference values. In the on-line phase, mean filtering is used to reduce the influence of the single random error of the signal.Results show that the improved KNN-PIT algorithm can better estimate the user's actual location,and decrease significantly localization errors,and improve real-time of localization.