摘要
表情图像是基于表情识别技术分析学习情感的基础,然而现有表情库样本数量有限、表情图像中仅包含单人表情、拍摄场景为实验室环境,这些局限性无法支持学习情感分析的深入研究。为了解决这些问题,文章搭建了北京师范大学学习情感数据库(Beijing Normal University Learning Affect Database,BNU LAD),同时对情感数据的标注方法进行了深入研究,最终形成了包含144位学习者的22708张表情图像、1792组图像序列及243段视频片段的学习情感数据库。该数据库对学习环境下的学习情感分析具有重要意义。
As the basis of the affective states analysis based on facial expression in learning, expression database is of great important. However,the existing databases are hardly to be used in supporting the in-depth study of affective states analysis, because the number of samples was small, the images were collected under laboratory conditions, there was only one person in each image. In order to realize the affective recognition and analysis of the facial expression in the process of learning, Beijing Normal University Learning Affect Database(BNU LAD) is built. The method of attaching the affective label and expressional label to affective expressions is studied deeply in the process of building database. The database consists of 22708 facial images, 1792 groups of image sequence and 243 video clips of 144 undergraduate and postgraduate students. The database is helpful to the development of affective states recognition in learning environment.
出处
《现代教育技术》
CSSCI
2015年第10期99-105,共7页
Modern Educational Technology
基金
北京自然科学基金"三维虚拟教学环境中虚拟教师和教学过程建模的研究"(项目编号:4102030)
中央高校基本科研业务费专项资金"自发性课堂学习情感的视觉建模与计算"(项目编号:2014KJJCA15)
教育科学十二五规划课题"智能感知技术在中小学作业减负中的应用研究与实践探索"(项目编号:DCA140229)资助
关键词
表情数据库
学习情感识别
情感标注
表情标注
facial expression database
affective states recognition in learning
affective expression label
facial expression label
