関西学院大学 研究活動情報

Kwansei Gakuin University Research Activities

土方嘉徳・商学部教授が「Speech-Driven Facial Animation by LSTM-RNN for Communication Use」で国際会議の優秀論文賞を受賞
Professor Yoshinori Hijikata, of the School of Business Administration, received the Best Paper Runner-up Award at the Asia Pacific Workshop on Mixed and Augmented Reality (APMAR2019)

2019.06.07

受賞 Award

論文 Article

共同研究 Collaborative Research

 土方嘉徳・商学部教授が3月28、29日、奈良先端科学技術大学院大学で開催されたAsia Pacific Workshop on Mixed and Augmented Reality (APMAR2019)で、優秀論文賞(Best Paper Runner-up Award)を受賞しました。

 論文のタイトルは「Speech-Driven Facial Animation by LSTM-RNN for Communication Use」。この研究では、音声会話から3次元のCGキャラクタの顔表情を自動的に生成するシステムを開発しています。通常、CGキャラクタの顔表情を生成するには、人間の顔画像からリアルタイムに表情を検出して利用しますが、これではカメラという入力装置が必要なこと、顔領域の検出にはエラーが不可避であること、人は必ずしも顔表情を作り出せる環境にないことから、実用的ではありませんでした。この研究では、LSTM-RNNというディープラーニングの手法を用いて、音声会話のみで表情を自動的に生成できるシステムの開発に成功しました。

■受賞論文
「Speech-Driven Facial Animation by LSTM-RNN for Communication Use」
著者:Ryosuke Nishimura, Nobuchika Sakata, Tomu Tominaga, Yoshinori Hijikata, Kensuke Harada, Kiyoshi Kiyokawa
※同研究は、関西学院大学商学部、大阪大学大学院基礎工学研究科、奈良先端科学技術大学院大学先端科学技術研究科の共同研究です。

 Professor Yoshinori Hijikata, of the School of Business Administration, received the Best Paper Runner-up Award at the Asia Pacific Workshop on Mixed and Augmented Reality (APMAR2019), which was held at the Nara Institute of Science and Technology on March 28th and 29th.

  The title of the paper was “Speech-Driven Facial Animation by LSTM-RNN for Communication Use.” In this research project, Prof. Hijikata and his colleagues are developing a system that automatically generates the facial expressions of three-dimensional CG characters from voice communication. Typically, in order to generate facial expressions for CG characters, systems would detect and use human facial images in real-time. However, this was not practical, since a camera is necessary as an input unit, errors are inevitable for facial area detection, and people are not always in environments where they can produce facial expressions. This research project, which uses a deep learning method called LSTM-RNN, successfully developed a system that could automatically generate facial expressions through voice communication alone.
 
■Award-Winning Article
“Speech-Driven Facial Animation by LSTM-RNN for Communication Use”
Authors: Ryosuke Nishimura, Nobuchika Sakata, Tomu Tominaga, Yoshinori Hijikata, Kensuke Harada, Kiyoshi Kiyokawa
*This is a joint research project being conducted by the Kwansei Gakuin University School of Business Administration, the Osaka University Graduate School of Engineering Science, and the Nara Institute of Science and Technology Graduate School of Science and Technology.
 

戻る
back