An effective random statistical method for Indoor Positioning System using WiFi fingerprinting

摘要

In the paper, an effective random statistical method is proposed for Indoor Positioning System (IPS) using WiFi fingerprinting. The proposed method consists of two phases: the offline handling process and the online positioning process. The offline handling process is used to collect a large number of WiFi signals at each indoor reference point and then create an offline database. This process handles the noise of WiFi signals and normalizes the database about location fingerprints for IPS. To further improve the accuracy of indoor positioning, the Mahalanobis distance is utilized to determine the indoor location for the online positioning process. Compared to the Weighted K-Nearest Neighbor (WKNN) algorithm based on Euclidean distance, experimental results show that it can improve the positioning accuracy using the proposed random statistical method. For the proposed random statistical method, the maximum positioning error is less than 0.75 meters. However, the average positioning error is 1.5 meters using the WKNN algorithm. In addition, it can effectively handle the noise of WiFi signals using the proposed random statistical method in different indoor environments.