Gesture recognition through angle space

Abstract

As the notion of ubiquitous computing becomes a reality, the keyboard and mouse paradigm become less satisfactory as an input modality. The ability to interpret gestures can open another dimension in the user interface technology. In this paper, we present a novel approach for dynamic hand gesture modeling using neural networks. The results show high accuracy in detecting single and multiple gestures, which makes this a promising approach for gesture recognition from continuous input with undetermined boundaries. This method is independent of the input device and can be applied as a general back-end processor for gesture recognition systems.