Sign Language Recognition Of Deaf-Mute Persons Based On The Monocular Video

In recent years, with the development of computer technology, the technology based on sign language recognition is becoming a new development trend at the same time. The important core of this paper is composed of two parts, which are static alphabet sign language recognition and dynamic sign language recognition. The paper constructed a static sign language recognition system based on the character extraction algorithms of Fourier and K-Nearest Neighbor (KNN). Besides, it has constructed a dynamic sign language recognition system based on the tracing algorithm of continuously adaptive mean-shift(camshift) and the recognition algorithm of Dynamic Time Warping (DTW).These two systems are all based on the skin color segmentation of hand gestures. In the process of static sign language recognition it build a database. The effect of direction and phase change of feature extraction has decreased after the traditional Fourier algorithm is normalized.And the consequence wills more accuracy of gesture features. Next, it uses the weighted KNN algorithm to addition and adjusting the database of static sign language and build a multi-class classifiers to recognition sign language. In the process of dynamic sign language recognition, it uses camshift algorithm to capture gestures images, tracking the path of gestures and get the area of gesture. Then use DTW calculation the minimum distances between gestures get the recognition results.In this paper, the specific work based on deaf-mute persons will be described as follows:(1)At first, it described the development process and research status of sign language recognition at home and abroad. Then concluded the development trend of sign language recognition technology and determined the research direction of this paper.(2)After studying the methods of gesture positioning, it chooses a skin color template to achieve the segmentation location by building a better color template.(3)For static sign language, the Fourier algorithm is used to describe the feature extraction. The weighted KNN algorithm is used to classification and recognition the sign language to get results.(4)It uses camshift algorithm to capture gestures images, traeking the path of gestures and get the area of gesture. The above is the foundation of the dynamic sign language recognition. Then use the DTW algorithm in the part of dynamic sign language recognition after gesture feature extraction.(5)A large number of experiments are done after determine the algorithms of sign language recognition system. The feasibility of each algorithm is validated. The recognition rate of static alphabet sign language and dynamic sign language is verified.According to a plenty of experimental results and data, the feasibility of the system's overall frame is verified in this paper. The algorithms of gesture recognition system are able to identify similar sign.