http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
황대원(Daewon Hwang),정정민(Jungmin Chung),서봉원(Bongwon Suh),서경진(Kyoungchin Seo) 한국디지털콘텐츠학회 2022 한국디지털콘텐츠학회논문지 Vol.23 No.11
As speech and motion play a essential role in communication, studying co-speech gesture is a long-standing research topic. High-quality audio and gesture data are desired in both research and industry fields. However, only English-based audios are available and types of modality are also limited. In this study, we collect and propose 761 hours of Korean-based large-scale co-speech gesture dataset (KLSG) annotated with the multi-modal information, such as multi-view videos, script, emotions, etc. All corresponding data are synchronized with speech audio time. To verify the validity of KSLG dataset, we compared the results of various co-speech gesture generation tasks using KSLG and a benchmark dataset. The result shows that our dataset has similar characteristics with the benchmark dataset when training gesture generation models. A user study also suggests that our dataset contains Korean speech and behavior features.