人脸部分遮挡条件下表情识别研究的新进展  

New Research Advances in Facial Expression Recognition Under Partial Occlusion

作　　者：蒋斌 李南星 钟瑞 吴庆岗 常化文 JIANG Bin;LI Nanxing;ZHONG Rui;WU Qinggang;CHANG Huawen(College of Computer and Communication Engineering,Zhengzhou University of Light Industry,Zhengzhou 450001,China)

机构地区：[1]郑州轻工业大学计算机与通信工程学院,郑州450001

出　　处：《计算机工程与应用》2022年第12期12-24,共13页Computer Engineering and Applications

基　　金：国家自然科学基金(61702464);河南省高等学校重点科研项目(16A520028);郑州轻工业学院博士科研基金(2014BSJJ077);河南省科技攻关项目(222102210103,222102210039,192102210136,212400410154)。

摘　　要：随着人工智能技术的发展,人脸表情识别可以从图像或视频中抽取表情状态,识别对象心理情绪,从而达到更好的人机交互效果。然而多数方法只关注正面无遮挡的人脸表情识别,并不能适用于客观复杂的场景,极大地限制了算法的实用性。近几年,针对光照遮挡、噪声遮挡、姿态遮挡、实物遮挡等不同类型的遮挡,研究者们提出了各种新方法来挑战人脸部分遮挡下的表情识别,综述了这些方法的主要原理并进行了对比分析,同时对未来的研究和发展方向进行了展望。With the development of artificial intelligence technology,facial expression recognition can extract expression states from images or videos to recognize the mental emotions of human beings,achieving a better human-computer interaction effect.However,there are many methods only focusing on facial expression recognition without occlusion,which is not applicable to objective and complex scenarios and greatly limits the practicality of the algorithm.In recent years,aiming at different types of occlusion such as illumination occlusion,low resolution occlusion,pose variations,and physical occlusion,researchers have proposed various new methods to solve these problems.The survey introduces the core ideas of relevant methods,and conducts a comparison analysis about these methods,meanwhile looks forward to future research and development.

关 键 词：人脸表情识别 部分遮挡 新进展 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]