人脸情绪识别研究综述  

Survey of Facial Emotion Recognition

作　　者：宋佳 蔡峰权 顾天晴 曾清源 谭定英[1] 陈平平[1] SONG Jia;CAI Fengquan;CU Tianqing;ZENG Qingyuan;TAN Dingying;CHEN Pingping(College of Medical Infomation Engineering,Guangzhou University of Chinese Medicine,Guangzhou 510000)

机构地区：[1]广州中医药大学医学信息工程学院,广州510000

出　　处：《现代计算机》2021年第22期133-139,155,共8页Modern Computer

基　　金：广东省大学生创新创业训练计划项目(No.S202010572110)。

摘　　要：从人脸情绪识别技术的研究背景、数据集、识别方法、发展与挑战四个方面,归纳总结现阶段人脸情绪识别方法的研究进展与运用。在中国知网、万方数据以及X-MOL用"人脸情绪识别"、"CNN"、"深度学习"等关键词检索总计约306篇相关文献,并根据其内容相关度、权威信、表发时间等原则筛选出30篇。基于面部表情与语音信号等结合的多模态情绪识别已成为当前国内外研究热点。通过对各类文献综合分析,发现CK+数据集在各研究中的综合表现较为优异;人脸情绪识别技术的优化与应用,面临着资源、技术、伦理等诸多挑战。将进几年国内外文献分析结果提炼,为探索人脸情绪识别技术的研究人员提供了方案设计思路。From the four aspects of the research background,data set,recognition method,development and challenge of facial emotion recognition technology,summarize and summarize the current research progress and application of facial emotion recognition methods. A total of about306 related documents were retrieved from CNKI,Wanfang Data and X-MOL using keywords such as face emotion recognition,CNN,and deep learning,and based on their content relevance and authoritative information. Thirty articles were selected based on the principles of,time of publication,etc. Multi-modal emotion recognition based on the combination of facial expressions and voice signals has become a current research hotspot at home and abroad. Through a comprehensive analysis of various documents,it is found that the comprehensive performance of the CK + data set in various studies is relatively excellent;the optimization and application of facial emotion recognition technology faces many challenges such as resources,technology,and ethics. The results of the analysis of domestic and foreign documents in the past few years are refined to provide a solution design idea for researchers exploring facial emotion recognition technology.

关 键 词：人脸情绪识别 深度学习 卷积神经网络 人脸表情数据集 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]