Feature Analysis in Indoor Positioning

In order to satisfy the higher precision of indoor location-based service (ILBS), scholars have explored a great deal of algorithms based on Wi-Fi, ultrasonic, RFID or infrared, but all of which need additional device settings for transmitting and receiving signals before implementing location recognition. This paper proposed an idea that how to conveniently find the optimal feature or composite features for more accurate indoor positioning, which were achieved by classification and clustering algorithms integrated in WEKA software. All the samples on five features were collected only with the help of a smart-phone lasted for 15 days. Comprehensive experiments, comparative studies with different kinds of samples, and the correlative performance evaluations were also completed. The results proved that our proposed schema was rational: the optimal feature, combinatory features and the corresponding statistical index of samples can be selected by classification and clustering for location recognition.