An Individual Behavioral Pattern to Provide Ubiquitous Service in Intelligent Space

摘要

We develop an intelligent space "Tagged World", aiming to provide proactive services according to user intention in a ubiquitous environment. User behavior must be precisely recognized to provide appropriate services for individual user. This paper proposes a user behavior detection method with a personalized behavioral pattern in the Tagged World which identifies objects a user touches. They are adapted to individual user. Our method focuses on special scenes in which user's mode significantly changes, such as a scene of going out. Proactive services can be provided to user most effectively in such scenes. The method detects user behavior with a behavioral pattern represented by discrete order of target objects of user operation. It separates the order check from a probabilistic model. Because a behavioral pattern can be adapted to individual user in a short time, the method can start providing services to user early. Experiments have proved that our method detects more than 90% of user behavior correctly with a behavioral pattern which is created in a practical short time with less than 10 sample behavior logs.