An improved word segmentation algorithm for lip-reading

An improved word segmentation algorithm is proposed, which is used to find out the section that represents the pronunciation part from the video frame sequence in the computer lip-reading system. In the computer lip-reading system, after extracting the lip features, the standard deviation of the lip features is calculated. Then the paper proposes the word segmentation method based on the standard deviation and the results show that this method can only be used to determine the approximate pronunciation section and that cannot accurately determine the location of the beginning and the ending of the pronunciation position. So an improved method is presented that the location of the beginning and the ending of the pronunciation section is determined according to the distribution of the gradient of the standard deviation. The results show that the improved method can greatly improves the accuracy of locating of the beginning and the ending of the pronunciation section.