Recognition of Dynamic Hand Gestures with Time-of-Flight Cameras

In this paper an algorithm to recognize single stroke gestures using a Time-of-Flight camera is presented. To achieve this goal, freehand gesture input is provided by articulating a single stroke in three-dimensional space using a stretched nger. The position of the ngertip is tracked and gesture input can be deactivated by forming a st. In this way, the start and end of a gesture is recognized automatically. The gesture is then recognized with the $1 gesture recognizer which is adapted to depth data. The algorithm is applied to recognizing ten gestures composed of the numbers zero to nine and evaluation results reveal a ngertip tracking accuracy of over 88% and an average gesture recognition rate of 75%.