Preliminary Study of Classifier Fusion Based Indoor Positioning Method

Abstract

Indoor positioning technology is commercially available now, however, the positioning accuracy is not sufficient in the current technologies. Currently available indoor positioning technologies differ in terms of accuracy, costs and effort, but have improved quickly in the last couple of years. It has been actively conducted research for estimating indoor location using RSSI (Received Signal Strength Indicator) level of Wi-Fi access points or BLE (Bluetooth Low Energy) tags. WiFi signal is commonly used for the indoor positioning technology. However, It requires an external power source, more setup costs and expensive. BLE is inexpensive, small, have a long battery life and do not require an external energy source. Therefore, by adding some BLE tags we might be able to enhance the accuracy inexpensive way. In this paper, we propose a new type of indoor positioning method based on WiFi-BLE fusion with Fingerprinting method. WiFi RSSI and BLE RSSI are separately processed each one by a Naive Bayes Classifier. Then, Multilayer Perceptron(MLP) is used as the fusion classifier. Preliminary experimental result shows 2.55m error in case of the MLP output. Since the result is not as good as the ones using conventional method, further test and investigation needs to be performed.