Kinect-based gesture recognition and its application in MOOCs recording system

Gesture recognition is increasingly remarkable in the field of HCI, since hand motions and gestures enable users to interact with computers in more natural ways. This paper focuses on two-hand gesture recognition and proposes a Kinect-based method which specially takes a certain static gesture as a start and end mark of a dynamic gesture. Furthermore, we use an innovative way to extract the feature of gesture and then build the Coupled Hidden Markov Model for training and recognition. The experiment results indicate that the gesture recognition algorithm is efficient and accurate. On the basis of this method, we present a brand-new Kinect-based MOOCs recording system which allows teachers to record MOOC(Massive Online Open Course) videos with PPT operations simply on their own and finally outputs two videos (screencast and camera video) with the same timeline so that extra postproduction to synchronize the two video sources can be avoided. And it is proved that the system can meet almost all needs in MOOCs recording and teaching practice.