The Algorithm Of The Robot's Body Gesture Gesture Recognition And Imitation

Human-Robot Interaction(HRI) refers to the interaction and communication betweenhuman and robot. The ultimate purpose is to make the robot be able to understand humanlanguage and behavior, so that human can communicate with robot more effectively andnaturally. In order to communicating naturally, we must refuse the traditional input-outputmode and make the HRI be similar to communication between people. Thecommunication between people depends on the voice and vision heavily, so HRI will bealong the voice and visual interaction development. This paper proposes a HRI methodbased on human pose estimation.Firstly, the development of human-computer interaction and human-robot interactiontechnology and human pose estimation research results are introduced. Then, the humanpose estimation algorithm adopted here is described which includes the construction ofpart detector and multi-index evaluation method. At last, the experiments are completedon the NAO robot. On the basis of previous work, the main work is as follows:(1) Human pose recognition and action imitation are implemented by using a NAOrobot as test bench. Besides this, we realize that make robot move according toour special pose instruction.(2) A part detector for each human body part is obtained by AdaBoost algorithm.Those images are normalized before training and the inner class distance can bedecreased. By doing this, the performance of part detectors is improved. Thenormalized method is to rotate the image patch until its spindle is vertical. Whenwe use these part detectors, we must rotate the tested image in order to find all thecandidate regions.(3) Then, these candidate regions are evaluated by using several defined evaluationindexes in order to select an optimal one. The evaluation indexes include partconformity, foreground coverage rate, foreground coverage entropy, priorprobability, correlation and skeleton density and intermediate level. (4) Some primary researches are done about the spatial consistency and temporalconsistency of NAO robot.