LITEVSR: Efficient Visual Speech Recognition by Learning from Speech Representations of Unlabeled Data | IEEE Conference Publication | IEEE Xplore