IEMOCAP（the Interactive Emotional Dyadic Motion Capture）

交互式情感二元运动捕捉 (IEMOCAP) 数据库是一个动作，多模式和多扬声器数据库，最近在南加州大学的SAIL实验室收集。它包含大约12小时的视听数据，包括视频，语音，面部动作捕捉，文本转录。它由二元会议组成，演员在其中表演即兴创作或脚本场景，这些场景是专门为引起情感表达而选择的。IEMOCAP数据库由多个注释者注释为分类标签，例如愤怒，幸福，悲伤，中立，以及价，激活和优势等维度标签。详细的运动捕捉信息，引发真实情感的交互式设置以及数据库的大小使该语料库成为社区中现有数据库的宝贵补充，用于研究和建模多模式和表现性人类交流。

The Interactive Emotional Dyadic Motion Capture (IEMOCAP) database is an action-oriented, multimodal and multi-speaker corpus recently collected at the SAIL Lab of the University of Southern California (USC). It contains approximately 12 hours of audiovisual data, including video recordings, speech, facial motion capture data and text transcriptions. The database consists of dyadic interactive sessions where actors perform improvised or scripted scenes specifically selected to elicit emotional expressions. The IEMOCAP corpus is annotated by multiple annotators with categorical labels such as anger, happiness, sadness, neutral, as well as dimensional labels including valence, activation and dominance. The detailed motion capture information, the interactive setting that elicits genuine emotions, and the size of this corpus make it a valuable addition to existing corpora in the research community for research and modeling of multimodal and expressive human communication.

搜集汇总

背景与挑战

背景概述

IEMOCAP是一个由南加州大学于2008年发布的多模态情感数据库，包含约12小时的视听数据，涵盖视频、语音、面部动作捕捉和文本转录。该数据集通过二元会议中的即兴或脚本表演收集，旨在引发真实情感表达，并标注了分类和维度情感标签，适用于多模态情感交流研究。