An self-adaptive wireless indoor localization system for device diversity

In this paper, we proposed a self-adaptive wireless indoor localization system for device diversity. The major feature is that our system can incrementally collect RSS data for device calibration without stopping the localization service. The system has two phases. In the localization phase, we used relative radio signal strength (RSS) features as inputs to overcome device diversity and train classifiers for target localization. When a new user starts to use the system, this phase could provide coarse localization even if the device is different. After collecting more RSS data from user's device, in the calibration phase, our system aims to calibrate device difference adaptively. We proposed using a histogram-based method for RSS calibration. Compared with the previous calibration methods, neither manual pair-wise data collection nor the prior target device information is necessary in our system. After we determined the model for RSS calibration, all the new coming RSSs could be calibrated. By inputting the calibrated RSSs into the localization phase, the fine localization is achieved.