A dialogue interface for investigating human activities in surveillance videos

Abstract

In this paper we present a dialogue interface for investigating human activities in surveillance videos. The interface exploits the information computed by the recognition system to support users in the investigation process. The interaction dialogue is supported by a sketch language enabling users to easily specify various kinds of questions about both actions and states, and the nature of the response one wishes. The contribution of this research is twofold (1) proposing an intuitive interaction mechanism for surveillance video investigation, and (2) proposing a novel question-answering model to support users during the information-seeking process. © Springer-Verlag 2011.