Research On The Methods Of Lip Detecting And Tracking Based On Video Image

With developing of computer technology, image processing and pattern recognition technology, lip reading technology which has become a major research will be applied to speech recognition, intelligent human-computer interface as well as the deaf-assisted education and other fields. In this paper, we described the lip of detecting and tracking methods which are main technology of lip-reading system based on the video sequences, the main work is listed here:The automatic face detection is presented for the first frame video. We used color space conversion from RGB to YCbCr, established Gaussian skin color model according to the hue clustering characteristics for face detection.The methods of detecting and locating lip based on the combination of the geometric features and the hue separation characterist in YIQ space used the adaptive filter model. Compared with and the Red Exclusion algorithm, extraction of lip contour can be more accurate and effective.We used the deformable template for extracting geometrical features, obtained the initial parameters of the lip template, completed the template matching process according to the cost function minimizing. The methods of extracting and tracking lip features have advantages of the lower dimension, moreinformation in natural conditions over other available methods.