Kinect-based Human Posture Recognition And Correction

Human-computer interaction identify human action, such as gestures, body potential, facial expressions, to achieve interaction between human and machine, even combined with voice recognition to achieve multi-channel integration of human and computer interaction techniques. Human posture recognition is a key technology for the development of Human-computer interaction. People by making a variety of posture to pass information to the computer and the computer achieve an accurate understanding of the meaning of human posture by identifying the human body’s movements. Thus, for this natural harmony of human-computer interaction needs, human posture recognition technology has become a hot topic.We research on the algorithm of human body detection and tracking, use the depth image which obtained by Kinect to track the human body. In order to improve the accuracy of posture recognition, this paper presents a method of measuring human joints angles to identify human posture based on Kinect. This method is to use Kinect to get the space coordinates of body joints, to realize posture recognition by matching the angle of posture recognition; it can accurately recognize human posture in real time.We investigate the repair algorithm of human body joints for the human body parts are obscured by objects and the confusion between body parts when tracking. This repair algorithm combines with the invariance of the human skeleton length, the continuity of human movement and the rotation angle limit of joints to assess the trustworthiness of joints, and then use the assessed value to repair the joints. The joint which has been identified, we test its reliability, combined with behavior, reliability, and track the status of the kinematics and assign weights to them to determine the reliability of this joint. Combined joints repair algorithm and key points of posture correction method for the synthesis and optimization of human posture. After, we study the posture correction method based on joints, and synthesis and optimization the human posture.We design a system of human posture recognition and assessment based on Kinect. This system utilize depth image obtained by Kinect to get the skeleton joints, the function of angle measure achieved by the method of measuring the angle between the joints, and achieve the posture recognition by the information of these angle. In the process of assessment, if the user’s posture does not meet posture set condition, posture error message are prompted the user to be re-started. System eventually realizes the real-time recognition and assessment of human posture.