Abstract. Automatic lipreading is automatic speech recognition that uses only visual information. The relevant data in a video signal is isolated and features are extracted from it. From a sequence of feature vectors, where every vector represents one video image, a sequence of higher level semantic elements is formed. These semantic elements are "visemes" the visual equivalent of "phonemes" The developed prototype uses a Time Delayed Neural Network to classify the visemes.

Within the Helium3D project, a wide spectrum of gesture tracking aspects is investigated: Near-field (3D) versus Far-field (2D), proprietary sensors versus web-cams, discrete versus continuous gestures, and data modeling to cover all types of gestures and use cases. This paper gives an overview of those aspects. A common aspect of the described tracking technologies is that they don't rely on cues for which it is likely to encounter exceptions in the real-world, like skin-color filtering in image capturing. The event model generalizes trackers and applications, while at the same time offering a way of tracking persons while they move in or out of the scene

No abstract

No abstract