Emotion estimation by EEG using features from multidimensional-directed coherence analysis

カテゴリ: 部門大会

論文No: SS1-4

グループ名: 【C】2023年電気学会電子・情報・システム部門大会

発行日: 2023/08/23

タイトル(英語): Emotion estimation by EEG using features from multidimensional-directed coherence analysis

著者名: 鳥居 暖華（東京電機大学）,島田 尊正（東京電機大学）,阪田 治（東京理科大学）,深見 忠典（山形大学）

著者名(英語): Haruka Torii (Tokyo Denki University),Takamasa Shimada (Tokyo Denki University),Osamu Sakata (Tokyo University of Science),Tadanori Fukami (Yamagata University)

キーワード: 多次元有向コヒーレンス|脳波|情動推定情動推定|Multidimentional-directed coherence|Brain ativity|Emotion estimation

要約(日本語): In recent years, much research for estimating emotions with bio signals were conducted and reported. In our work, we tried to estimate four emotions by adopting several signal processing methods and neural network to EEG. In our experiment, first, we collected EEG data for four emotions, joy, sadness, anger and surprise which were induced by showing subjects emotional face images. Next, we extracted the features for estimating emotions by using the multi-dimensional directed coherence method. After that, we sorted the data in accordance with the rule that data has higher correlation lies closer. Finally, we estimated emotions by using neural network. As a result, the accuracy rate of each emotion may be higher than conventional method which uses manually-made classification rules.