Utilizing an accelerometric bracelet for ubiquitous gesture-based interaction

In this paper we present an approach for recognizing free-handed gestures using an embedded wireless accelerometric bracelet. We developed a very low complexity algorithm which can be directly implemented on the device and operate in real-time. New gestures can be easily added through supervised learning. An evaluation shows the feasibility of our approach. Simple gestures are detected and recognized at a very high rate (> 97%) while more complex ones were misclassified more often (48% - 95%).