Vision-based hand gesture interaction using particle filter, principle component analysis and transition network

Vision-based human-computer interaction is becoming important nowadays. It offers natural interaction with computers and frees users from mechanical interaction devices, which is favourable especially for wearable computers. This paper presents a human-computer interaction system based on a conventional webcam and hand gesture recognition. This interaction system works in real time and enables users to control a computer cursor with hand motions and gestures instead of a mouse. Five hand gestures are designed on behalf of five mouse operations: moving, left click, left-double click, right click and no-action. An algorithm based on Particle Filter is used for tracking the hand position. PCA-based feature selection is used for recognizing the hand gestures. A transition network is also employed for improving the accuracy and reliability of the interaction system. This interaction system shows good performance in the recognition and interaction test