Indoor positioning system based on BLE location fingerprinting with classification approach

Position estimation is an important technique for location-based services. Many services and applications, such as navigation assistance, surveillance of patients and social networking, have been developed based on users’ position. Although the GPS plays an important role in positioning systems, its signal strength is extremely weak inside buildings. Thus, other sensing devices are necessary to improve the accuracy of indoor localisation. In the past decade, researchers have developed a series of indoor positioning technologies based on the received signal strength (RSS) of WiFi, ZigBee or Bluetooth devices under the infrastructure of wireless sensor network for location estimation. We can compute the distance of the devices by measuring their RSS, but the correctness of the result is unsatisfactory because the radio signal interference is a considerable issue and the indoor radio propagation is too complicated to model. Using the location fingerprint to estimate a target position is a feasible strategy because the location fingerprint records the characteristics of the signals and the signal strength is related to the space relation. This type of algorithm estimates the location of a target by matching online measurements with the closest a-priori location fingerprints. The matching or classification algorithm is a key issue in the correctness of location fingerprinting. In this paper, we propose an effective location fingerprinting algorithm based on the general and weighted k-nearest neighbour algorithms to estimate the position of the target node. The grid points are trained with an interval of 2 m, and the estimated position error is about 1.8 m. Thus, the proposed method is low computation consumption, and with an acceptable accuracy.