The Research On Applying The Sign Language Recognition Technology For Deaf Education Based On Kinect

Sign language is the most important communication tool between deaf people. Sign language recognition technology is to make computer understands sign language and interprets it in an understandable way to make healthy people and deaf people communicate more conveniently and quickly. With the development of motion-sensor technology, makes computer more easier to understand motions and provides a new way for researching sign language technology.Traditional input tools can’t meet the requirements of human-computer interaction. So researching sign language recognition technology can not only makes deaf-mute people easier to communicate with healthy people, but also contributes to the development of human-computer interaction technology.This thesis based on the requirement of deaf-mutes school teaching, researches the traditional methods of sign language recognition.Using Kinect motion-sensing camera to improve the recognition technology and implements a sign language recognition system.This system includes dynamic sign language recognition part and static sign language recognition part.For dynamic sign language recognition part, it contains palm moving recognition and finger moving recognition. Using sign language recognition system makes deaf students get better access to various educational resources.And it lays the foundation of improving the interaction between teachers and computer.First this thesis introduces sign language recognition’s background and meaning and describes the status at home and abroad. Next it introduces common feature extraction algorithms and recognition algorithms.And then it presents some technologies and platforms and illustrates theoretical supports of using sign language recognition system to deaf-mutes teaching.Then detail described the method of hands’feature extraction and detection. In dynamic sign language feature extraction part, this paper according to Kinect characteristics proposes the methods of Hand Motion Twice Feature Extraction and Finger Motion Depth Histogram.In static sign language feature extraction part, this thesis uses Improved Freeman Chain Code. For hand motion recognition this paper presents Two Layer HMM Model. For finger motion recognition part this paper uses SVM/HMM Model,and for static sign language recognition part this paper uses SVM Model.At last, this paper validates the methods above and presents the design and implement of the system.