The Design Of Sign Language Recognition System

With the development of Human-Machine Interaction technology,somatosensory interaction has gradually become a research hotspot for scholars,and it has broad prospects in robotic remote control,leisure entertainment,gesture recognition and other application fields.As a special gesture the sign language is the main way for the deaf and mute people to express their ideas.However,most normal people do not understand the sign language therefore cannot communicate.The technology of sign language recognition can transform the actions into text or the other information as well as create a barrier-free communication platform between deaf and dumb people,which has great significance for research.Kinect camera is selected in this paper as a data acquisition equipment and uses the color image and bone data information mainly which provided by it to recognize the sign language in real time.The static sign language and the dynamic isolated sign language are identified mainly in this paper.The details of this paper as follows:1.In the static sign language recognition section:the algorithm of smoothing,skin color segmentation,binarization,and morphologically processing are used on the obtained color image in the stage of data preprocessing.Hu moments is used to match the extracted hand area image with the template image during the identification phase.A static sign language recognition system is achieved which can recognize the sign language figures from 1 to 9.2.In the dynamic sign language recognition section,the bone data is normalized to reduce the errors of body shape differences and relative distance.First Hidden Markov Model algorithm is used to train data,but the model training is not perfect because of the amount of data is small.Instead,Dynamic Time Warping and k-Nearest Neighbor classification algorithm are used to classify and recognize dynamic sign language.3.A system has designed and implemented in Visual Studio that can recognize the sign language in real time based on the above algorithms in this paper.The system can realize the sign language recognition on real-time for non-specific people.The data base is used in this system for accuracy testing and experiments show that the system has a good recognition rate.