Algorithm Study About Lip Figure Extraction And Recognition Based On Image Processing Technology

The biometric identification technology has developed gradually with the rapid development of the computer technology. As a biometric identification technology, lip-reading is not only used in traditional deaf communications, intelligence access, but also widely used in the field of modern identity, and has become a hot research and application both at home and abroad.Lip-reading research includes: the basic characteristics of the mouth figure extraction, description and said, as well as lip figure recognition and understanding, it's the comprehensive understanding of the relevant information from number of channels.The characteristics split and edge extraction from the lip figures focus on this paper. Based on analysis and compares of the current mainstream method of lip-reading deeply, the characteristics of the lips separated by combining of neural network and gray threshold. After several training and select the appropriate threshold values according experience, a better lip segmentation binary image will be obtained. And then smoothing the image, extracting the edge of the image by structure of the template, and selecting amount of appropriate points from the edge features; Finally, a smoothing fitting curves will be obtained by training of the edge feature using of neural networks.A simple library of Chinese vowel Lamiaceae is established in this paper to carry out more accurate recognition. Using the GUI of the MATLAB to make a user interface for lip-reading system which can preview video images, select image, separate and understanding lip, etc.