An Improved Fingerprint Method for Indoor Mobile Object Positioning

The mobility of people in big cities tends to increase with the improvement in the availability of sharing access in public facilities. However, the dense environment with moving objects will allow the possibility of the object's detachment from a monitoring system. This condition will cause concern, especially if the object is a priority that needs to be protected. We propose a system for detecting moving objects in indoor environments using a fingerprint method of Received Signal Strength (RSSI) data retrieval. The work was conducted in the observation area, a part of a mall full of tenant stores and humans moving during shopping hours. The randomness of RSSI data emitted from access points in the indoor area is affected by multipath and signal reflection because of walls or furniture existing in the building. The K-NN regression algorithm was utilized to generate RSSI data based on the on-site measurement to prune this randomness. The generated data will be clustered, using the K-means algorithm and Elbow method to ensure the optimal number of K. The experiment results showed that using the best K for RSSI clustering, the positioning accuracy produced by the system reached 77% of the total expected accuracy. Meanwhile, according to the signal characteristics of indoor buildings, the entrance corner had the worst data distribution, with 37.35% of the data generated having an RSSI value close to the receiver sensitivity threshold.