Anticipatory Matching Method for Query-Based Head Gesture Identification

This paper presents a matching method for head gesture identification in a query-based interface. Head gesture identification has been achieved by various methods ; DP matching ; hidden Markov model, and so on. The aims of these conventional methods are dealing with general gestures at arbitrary moment. On the contrary, in our purpose, the timing of a gesture or what gesture will be made are predictable to some extent. The main purpose of this research is to realize an interface under variety of conditions, in flexible situations every time, and everywhere, for example, during walking around. In our system, there are two levels of anticipation features. The one is modifying the reference patterns according to the context of the query and cognitive difficulty of the question before making query. And the other is incursive refinement of reference patterns during matching process with allowance of future revision.