Monitoring and predicting human-place interaction for mobile context-aware applications

Abstract

As location-aware mobile devices such as smartphones are prevalent these days, location-based applications become very popular and ubiquitous in our daily lives. Introduced in diverse domains such as social networks (e.g., Foursquare and Google Latitude), transportation (e.g., Google Navigation), commerce (e.g., Google AdMob and Shopkick), etc., location-based applications provide users with more convenience than ever before. However, the majority of existing location-based applications use very limited information such as current location. For example, they recommend nearby restaurants based on a user’s current location. Such static and limited information is insufficient to provide not only new functionality but also rich user experiences.

Beyond simply using a user’s current location information, understanding and exploiting human-place interaction gives more opportunities to provide true mobile context-aware applications. People usually use diverse places such as home, office, restaurants, etc. in their daily lives. In other words, a person’s daily life can be considered as a sequence consisting of interaction with diverse places. Therefore, if we can comprehensively understand human-place interaction, we can provide rich user experiences that can eventually enrich their daily lives. For example, we may imagine that a service recommends some nice places by predicting places where a user would highly likely visit next.

From this perspective, in this dissertation, we provide novel systems enabling comprehensive human-place interaction monitoring, and novel applications using the systems. To comprehensively understand hu-man-place interaction, we divide the problem into two sub-problems: in-building human-place interaction and city-wide human-place interaction. In each problem, we discover unique challenges and provide interesting approaches.

First, we provide a smartphone system (i.e., VisitSense) enabling in-building human-place interaction monitoring ...