SEWA DB

该数据库包含来自六种文化 (50% 女性) 的398人的2000分钟以上的视听数据，并且年龄范围统一为18至65岁。在两种不同的环境中记录受试者: 观看广告和在视频聊天中讨论广告。该数据库包括面部标志、面部动作单位 (FAU) 、各种发声、镜像以及持续重视的效价、唤醒、喜好、一致、和 (不) 喜欢的原型示例。该数据库旨在成为情感计算和自动人类感知研究人员的非常有价值的资源，并有望推动人类行为分析的研究，包括文化研究。

This database encompasses over 2,000 minutes of audiovisual data collected from 398 participants across six cultural groups, with 50% identifying as female, and a standardized age range of 18 to 65 years. Participants were recorded in two distinct experimental settings: watching advertisements and discussing advertisements via video chat. The database includes facial landmarks, Facial Action Units (FAU), various vocalizations, mirroring behaviors, as well as continuously annotated valence, arousal, liking, agreement ratings, and prototypical examples of (dis)liking. This database is intended to serve as a highly valuable resource for researchers in affective computing and automatic human perception, and is poised to advance research on human behavioral analysis, including cultural studies.

AI搜集汇总

背景与挑战

背景概述

SEWA DB是一个多文化视听情感数据库，包含398名受试者在观看广告和视频聊天环境下的2000多分钟数据，标注了面部动作、发声和情感维度如效价与唤醒。该数据集由伦敦帝国理工学院于2019年发布，旨在为情感计算和人类行为分析研究提供资源，特别适用于文化差异研究。