Unsupervised indoor localization with motion detection

Abstract

Unsupervised indoor localization has received increasing attention in recent years. It enables automatically learning and recognizing the significant locations from Wi-Fi measurements continuously collected from mobile devices in a user’s daily life, without requiring data annotation from professional staff or users. However, such systems suffer from continuous Wi-Fi collection, which results in a high power consumption of mobile devices. These problems can be addressed through activating Wi-Fi collection when it is necessary and deactivating Wi-Fi collection when “enough” data is collected. By using the acceleration readings from the embedded accelerometer sensor, a motion detection algorithm is implemented for an unsupervised localization system DCCLA (Density-based Clustering Combined Localization Algorithm). The information of motion states (i.e. a mobile device in motion or not in motion) is then used to automatically activate and deactivate the process of Wi-Fi collection, and thus save power. Tests carried out by different users in real-world scenarios show an improved performance of unsupervised indoor localization, in terms of location accuracy and power consumption.