Improving location fingerprinting through time-shifted interval labeling

Abstract

Indoor localization and positioning systems have become increasingly important for many applications in ubiquitous computing. Previous work in this area has shown that Fingerprinting is a feasible and convenient technique for realizing indoor localization systems using WiFi radio signals. However, these systems require a time-consuming and costly training phase to build the radio map. Moreover, because maintaining the radio map requires great effort, such systems perform poorly in environments where radio signals change rapidly and fluctuation is high. We introduce a new concept called ‘time-shifted interval labeling’ that solves these problems. By using an accelerometer to detect whether a device is moving or still, the system can learn from all radio measurements in a stable time interval whenever it was previously labeled by the user or once the user retrospectively assigns a label to that interval.