Research On Static Sign Language Recognition Technology In Complex Background

Since twenty-first Century,human-computer interaction system has become the darling of the scientific community and has been favored by a large number of research scholars,in which the study of sign language recognition is very popular.With the deepening of the research and the improvement of social needs,the requirement of sign language recognition is more demanding.Vision based sign language recognition has become the focus of current research.The traditional sign language recognition method has a higher requirement on the environmental background and the image acquisition device and the human gestures.Therefore,in view of the complexity of the sign language image background and the similarity of the sign language shape as well as the limitation of the sign language recognition algorithm,this paper proposed a new static sign language recognition system.Due to the complexity of the background of sign language image,this paper proposed a new segmentation algorithm based on the improved skin color clustering and the graph cut method.By this method can obtain the complete hand region in complex background.Secondly,aiming at the similarity of hand shape,the method of chain code tracking is used to detect the contour of hand signed image.Then,the convex defects,the Hu moments,the geometric features and the regional descriptors are extracted as multiple features to classify gestures.The multiple features are comprehensive and discriminative,which can well reflect the characteristics of each sign language.Finally,due to the diversity of sign language and the limitations of sign language recognition algorithm,utilizing the SVM to classification that consists of the first classification on the number of defects and the second classification through multi-feature fusion.Later,for some difficult to distinguish and very similar sign language,two regions are used for fine classification.It can be better to classify each sign language by classification and recognition of sign language images more than one.The average recognition rate of 26 kinds of static sign language is 93.18% in self collected image library.The results show that the system designed in this paper is effective and practical.