Research On Human-computer Interaction Technology In Immersive Virtual Reality

Virtual reality(VR) is a computer simulation of three-dimensional environment, which goal is to get a strong immersive feeling of being in a virtual environment. Then it requires a variety of technical collaboration, among which input devices are important tools. Currently, although no-touch interaction has been extensively studied, few associated with Oculus Rift has been occurred. Gesture interaction takes the hands into the virtual environment, and becomes a hot topic input device of VR for its nature and direct characteristics.Based on the research summary of skeletal animation technology, visual-based gesture recognition, and the analysis of existing problems about the interactive mode in immersive virtual reality, the main work of this thesis are carried out on gesture-based immersive virtual reality technology given in the following.Firstly, as most of the traditional gesture recognition method can't adapt to the actual conditions of real-time continuous gesture data stream, based on joint vector and naive Bayes gesture recognition algorithm, an efficient and accurate gesture recognition method is proposed. The proposed algorithm work on the skeleton information and joint motion vector feature representation, and introduces an adaptive naive Bayes algorithm. So it uses fewer samples to complete the training. Experimental results show that the proposed algorithm has a higher the recognition rate and recognition speed. Then it can effectively adapt to the continuous data stream, has a better compatibility for VR applications and significantly increases the user's sense of immersion in the virtual scenes.Secondly, as there are little interactive visual feedback scheme for VR, a 3D interface and 3D interaction methods are proposed based on the Skeletal animation and collision detection. Experiments show that the proposed method can effectively enhance the immersive experience through gestures.Finally, by integrating above gesture recognition and collision detection technologies, an experimental platform for virtual reality interaction is designed and developed. This system contains a sample scenario applied with all the proposed methods.