There are a certain number of arm dysfunction patients whose legs could move. Considering the neuronal coupling between arms and legs during locomotion, this paper proposes a novel human-robot cooperative method for upper extremity rehabilitation. Legs motion is considered at the passive rehabilitation training of disabled arm, and its traversed trajectory is represented by the patient trunk motion. A Kinect based vision module, two computers and a WAM robot construct the human-robot cooperative upper extremity rehabilitation system. The vision module is employed to track the position of the subject trunk in horizontal; the WAM robot is used to guide the arm of post-stroke patient to do passive training with the predefined trajectory, and meanwhile the robot follows the patient trunk movement which is tracked by Kinect in real-time. A hierarchical fuzzy control strategy is proposed to improve the position tracking performance and stability of the system, which consists of an external fuzzy dynamic interpolation strategy and an internal fuzzy PD position controller. Four experiments were conducted to test the proposed method and strategy. The experimental results show that the patient felt more natural and comfortable when the human-robot cooperative method was applied; the subject could walk as he/she wished in the visual range of Kinect. The hierarchical fuzzy control strategy performed well in the experiments. This indicates the high potential of the proposed human-robot cooperative method for upper extremity rehabilitation.