Research And Application Of The Human-computer Interaction Technology Based On The Computer Vision

Perceptual Interfaces is the main tide of the human-machine interaction technology development, the vision based interfaces technology become more and more important and many people realize this fact. The key in the VBI technology is the fast and correct recognition of human motion. So difficult that the human motion analysis it is, there are many theory and technology problems to be solved.In this paper, we study and research the human motion recognition and understanding in 2D space. The main areas we study include the pre-process of the input video frames, human body geometric and motional description and the recognition of some simple motion. We have the following main research content and innovations.First, we present a new skin color filter. In the skin color detection method based on the color space, the division value is variable with the illumination condition, To overcome these shortcomings, we present a new skin color filter based on the fussy model and the B-P neural network. .This method has a fast study speed and can work well in variable illumination condition. It could be used in the hand gesture recognition and the face detection in the complex surrounding.Second, we present a new characteristic point automatic mark method based on colorful point to solve the fast motion tracking problem. We also put forward a new background modeling method account for the problem caused by the surrounding change. Based on the knowledge that the human skeleton model could not go beyond the area of the body silhouette, we bring forward a quick extraction of human skeleton model method, and the experiment show that it is fast and efficient.Third, the three-level motion description method we put forward makes us have a good basis to acquire the fast, correct and accurate motion recognition. We also introduce two kinds of model recognition methods: the posture estimate method based on the region and edge and the method based on the hidden Markov Models. We present a new motion recognition method based on the three-level motion description method and HMM, and our experiment testified its feasibility.Finally, we design and implement a infant education game system. We analyzed the background of this system. We also introduce necessary three phases to build this system and the relate technology include ogre engine and OpenCV. Finally we show some effect picture of the system.