Abstract
Hand posture estimation is important to realize a high accurate gesture interface and has been studied extensively. However, most of conventional gesture recognition systems require high computational cost to estimate hand posture. This paper proposes a new model-based method. First, we estimate the joint angles based on hand model in the method. After that, we construct a reduced-order space for the measured angle by using PCA(Principal component analysis), which is used to recognize the hand posture. Then, our proposed method makes it possible to reduce the computational cost and noise. We confirmed that our experimental system can recognize gestures in real-time and be applied for robot manipulation.
