Lip-Reading using Neural Networks?

Lip-Reading has been practiced over centuries for teaching deaf and dumb to speak and communicate effectively with the other people. In this study, the use of neural networks in lip reading is explored. We convert the video of the subject speaking different words into images and then images are further selected manually for processing. As per the research the horizontal and the vertical distance between the lips varies for each and every word considering the close proximities of similar sounding words. Based on this research we can create the database of commonly used words and our neural network model can form clusters of words based on its intelligent approach. This approach can be associated with various voice recognition softwares and help in increasing their efficiency readily even in a noisy environment, and creating new dimensions for human computer interaction.