CN-CVS: A Mandarin Audio-Visual Dataset for Large Vocabulary Continuous Visual to Speech Synthesis | IEEE Conference Publication | IEEE Xplore