Research Of Hand Gesture Recognition With Complex Background Based On Vision

Gesture recognition, as a sort of more advanced human-machine interactive technology, has received increasingly amount of attention from many relative researchers.In this paper, many kinds of gesture recognizing methods are introduced and the best fits of our system is chosen after a deep research about it. Then the practical experiments of training classifiers are described by graphic detail in the3rd part. After the theoretical and preoperational work, in the4th part, the work float and construction of our system are shown by diagrams. The5th part illustrates the tests and results of our system, with the6th part, summarize and outlook, following it.Main points of this research can be listed as below:Take haar-like feature as the classification parameter, which can correctly reflect the difference between the colors of adjacent regions. Integral image can thoroughly increase the efficiency of training. Use adaboost algorithm when training classifiers and get two gesture classifiers eventually.Research the influence of gesture samples to the classifier. Open gesture are better than close gesture considering about the positive rate and the false positive rate. Also, complex background and decent variety of gesture in positive samples have positive effect to the training.Design and realize a gesture controlled video player in order to demonstrate the effect and result of gesture recognizing. The player can recognize static and dynamic gesture at real time with complex background.Thus, the method developed by this paper can be used to control video player.