MORAVIA: A Video-Annotation System Supporting Gesture Recognition

Gestures and gesticulation play an important role in communication, particularly in public speech. We describe here the design, development and initial evaluation of MORAVIA (MOtion Recognition And VIdeo Annotation): a collaborative web application for (semi)automatic gesture annotation. MORAVIA was conceived as a support for the automatic evaluation of a speech based on non-verbal components, that is, as much as possible independent from the verbal content. We adopt an evaluation model, based on quality metrics related to gestures and provided by experts in the education and psychology domain. The final goal is to design and implement a system able to detect the gestures using a video camera and a depth camera, such as the Microsoft Kinect, to detect the position and the movements of the speaker. Then, the web application for video-annotation allows collaborative review and analysis of the dierent video sequences. This is useful both to domain experts, as a research tool, and to end users, for self-evaluation.