Dynamic Threshold Determination for Stable Behavior Detection

To provide services according to user behavior, parameters should be adapted appropriately for the precise recognition of user behavior. In particular, the threshold value which is used to create behavioral patterns matched for behavior recognition impacts accuracy of behavior recognition. Because the threshold value is common to all users in the conventional model, the threshold setting unsuitable for some users may cause low recognition rates. In this paper, we propose a behavior detection method which detects high-level user behaviors, such as “leaving home”. The proposed method achieves stable behavior recognition regardless of users, by introducing a model which dynamically determines the threshold value for individual user. Key–Words: Threshold, Context, Behavior, Ambient, Proactive