The Research Of Indoor Positioning And Tracking System Based On Signal Strength Using WLAN

With the development of information technology, location service is becoming an important requirement of people. However, the mature GPS(Globe Positioning System) can not be applied in indoor environment. So a new signal is needed for indoor positioning. Lately WLAN(Wireless Local Area Network) develops quickly, which is equipped in big public place. It makes attractable for WLAN to be used in indoor positioning. And the indoor positioning system using WLAN do not require any other hardwares and has more and more attentions of researchers. WLAN indoor positioning system becomes one of the most popular indoor positioning systems.This paper improves the classic WLAN indoor positioning system, and its contributions and innovation are:1. Classic indoor positioning system uses standard Gaussian distribution to model the signal strength received by an object in reference point. However, indoor environment is complicated and changing, existing multi-path interference and mobile people interference. Hence, this paper uses a multi-Gaussian mixture model to model the signal strength distribution in reference point. This model not only considers the influence of the interference, but also considers the related relationship between signal strength of every AP(Access Point), and improves the system performance. The parameters of this method are estimated by EM(Expectation Maximization) algorithm in this paper. The experimental results prove the effectiveness of our method.2. For indoor positioning system in a big environment, the number of reference points in fingerprint database is very big. Therefore, the clustering of fingerprint database is necessary. This paper mapped the elements of fingerprint database into a2-D plane using a self-organizing map. And the clustering is based on this2-D plane. The advantage of this method is that it can avoid the difficulty of clustering in space with high dimension, and the accuracy of clustering is high. The experimental results prove the effectiveness of our method.3. For tracking an object in indoor environment, the general method uses a Kalman filter. However, the noise of indoor environment is not Gaussian, for tracking an object it require an unscented Kalman filter or a particle filter. Unfortunately the particle filter algorithm is complex, and it is not comfortable for a real time system with a state whose dimension is more than3. This paper divides the object state into a2-D velocity state and a2-D location state. And it uses the unscented Kalman filter to estimate the velocity state and uses the particle filter to estimate the location state. This method not only improves the performance of the indoor positioning and tracking system, but also ensures that the system is real time. The experimental results prove the effectiveness of our method.