Research Of Mixed Hand Gesture Recognition Based On Video Image

Hand gesture recognition is one of the most popular researches in the field of human-computer interaction. It performs great efforts in constructing intelligent interaction. According to the defects of current researches on hand gesture recognition, this paper puts forward the conception of mixed hand gesture for building more natural and intelligent mode of interaction:(1)Including both static and dynamic hand gestures.(2)These two types of hand gestures can appear randomly.In order to recognize mixed hand gestures in video images, this paper performs deep research on these three aspects:hand segmentation, hand gesture division, hand gesture classification. First, we present an improved method which combines skin detection, foreground detection, and motion detection to realize hand segmentation in complex background environment. At the same time, we use the history information of hand position and the current information of motion to deal with the intelligent transformation of the operating hand. Then, as to the problem of hand gesture division, we suggest an algorithm based on hand states. The algorithm can effectively divide implicit static and dynamic hand gestures from the continuous image sequence. After that, we describe the methods for classifying static and dynamic hand gestures. With the purpose of recognizing complex dynamic hand gestures, we propose some new features to represent them and also a data aligning algorithm to solve the problem of inconformity among training data. At last, based on our research work, we develop a system for recognizing mixed hand gestures on line. And we show an example for using the system with a virtual environment system. The experimental results demonstrate that our methods can perform consequent, stable, and real-time recognition on natural mixed hand gestures.