Abstract: The properties of versatility and dexterity of the human hand have raised a growing interest both in Human Computer Interaction applications and in gesture recognition. Emphasis is laid on ease of detection and reconstruction of the hand posture as well as on real-time computation. This paper proposes a visual hand tracking system and a hand posture estimation method where the position and the orientation are recovered from interest cues on the hand. A kinematic model of the hand and a reconstruction method to recover hand posture are provided. A skeletal model of the hand is built to model the kinematic properties of the hand. A single camera provides the frame sequence of the mobile hand while color segmentation is used to detect salient features on the glove adorned hand. After the pose of the wrist is computed, the value of the finger joint angles are obtained by inverse kinematics. This method enables the successful recovery of the hand posture, opening the door to applications such as 3D input devices or powerful 6-dof control tools. Human hand skill can thus be exploited to control a robot gripper in a master-slave system taking advantage of the information provided by the human guide.