2021 17th International Conference on Machine Vision and Applications (MVA) | 2021
Japanese Sentence Dataset for Lip- reading
Abstract
This research is about lip-reading for Japanese sentences. Research on English sentences is actively pursued due to the extensive datasets. However, a sufficient dataset for Japanese sentences has not been released. Therefore, this paper builds a Japanese sentence dataset. A Transformer model is used for the recognition task. Three recognition target levels: phoneme, mora, and vowel, are set, and recognition experiments show that they can be recognized.