Using Binary Decision Tree and Multiclass SVM for Human Gesture Recognition

This paper presents a novel method to recognize the human gesture using binary decision tree and Multi-class Support Vector Machine (MCSVM). In a learning stage, 3D trajectory of the human gesture by a kinect sensor is assigned into the tree node of the binary decision tree according to its distribution property. The user's gesture trajectory is resampled and normalized, and we extract the chain code histogram at a regular interval. After training MCSVM in each node, we are able to recognize the human gestures.