Improved ROI and within frame discriminant features for lipreading | IEEE Conference Publication | IEEE Xplore