The goal of this project is the recognition of gestures which are recorded with a Wiimote or with a Kinect. The data are in case of Wiimote acceleration data and in case of Kinect 3D depth data.
Gestures from one or several persons are recorded and used for training. Later, gestures of the same kind from these persons or even from unseen persons shall be recognized. Our system reaches recognition rates of 90% and above.
We use learning methods like Slow Feature Analysis (SFA) or Hidden Markov Models (HMM).
