Skip to Main Content
Gesture recognition has great potential for mobile and wearable computing. Most papers in this area focus on classifying different gestures, but do not evaluate the distinctiveness of gestures in continuous recordings of gestures in daily life. This paper presents a new approach for the important and challenging problem of gesture recognition in continuous data streams. We use turning points of arm movements to identify segments of interest in the continuous data stream. The recognition algorithm considers both the direction of movements between turning points and the shape of the turning points for classification. Using the new method, seven gestures of different complexity are evaluated against a realistic background class of daily gestures in five different scenarios.