A gesture recognition apparatus using a multi-class SVM and a tree classification is disclosed. The three-dimensional information extraction unit extracts three-dimensional information on the user's joint based on the user's gesture input from the Kinect sensor. Gesture Classification Tree The design department designs gesture classification trees in advance according to the basic patterns of gestures, and classifies the extracted three-dimensional information according to the gesture classification tree. The chain code feature extraction unit extracts a feature vector of the chain code for the gesture. The gesture classifier classifies the gesture by learning with a multi-class SVM (Supprot Vector Machine) using the chain code and histogram information of the gesture as feature vectors. According to the present invention, a gesture can be correctly recognized through a natural user interface.
展开▼