K-EmoCon
收藏资源简介:
K-EmoCon是由韩国科学技术院创建的多模态传感器数据集,旨在支持自然对话中的连续情绪识别。该数据集包含来自32名参与者的多模态测量数据,包括视听记录、EEG和外围生理信号,这些数据是从大约10分钟长的16个配对辩论中使用现成设备收集的。与以往的数据集不同,K-EmoCon包含了所有三种可用视角的情绪标注:自我、辩论伙伴和外部观察者。评判者在观看辩论录像时,每隔5秒对情绪表现进行标注,涉及唤醒-价态和18种额外的分类情绪。K-EmoCon是首个公开可用的情绪数据集,支持社会互动中情绪的多视角评估。
K-EmoCon is a multimodal sensor dataset developed by the Korea Advanced Institute of Science and Technology to support continuous emotion recognition in natural conversations. It comprises multimodal measurement data from 32 participants, including audiovisual recordings, EEG signals, and peripheral physiological signals, which were collected using off-the-shelf devices across 16 paired debates each lasting approximately 10 minutes. Unlike previous datasets, K-EmoCon includes emotion annotations from all three available perspectives: the participant themselves, their debate partner, and external observers. Annotators label emotional expressions every 5 seconds while reviewing the debate videos, covering the arousal-valence dimensions and 18 additional categorical emotion categories. K-EmoCon is the first publicly available emotion dataset that supports multi-perspective assessment of emotions in social interactions.




