摘要
人脸表情识别技术是一个广泛的研究方向,涉及机器学习、图像处理、心理学等诸多领域,应用前景也相当广阔。近年来,随着人工智能等领域的发展和进步,属于情感计算领域的人脸表情识别技术也逐渐成为一个热门的研究方向。人脸表情识别任务一般由获取人脸图像、图像预处理、特征提取、特征分类4部分组成,人脸表情图像一般直接采用相关的数据集获取。首先介绍了人脸表情识别任务中需要进行的图像预处理步骤,以及特征提取和特征分类中的传统研究方法和深度学习方法,最后对人脸表情识别相关的数据集、发展趋势与挑战等进行阐述,并提出对未来的相关研究方法的看法。
Facial expression recognition is a research direction involving knowledge in many fields such as machine learning,image processing,psychology,etc.,and its application scenarios are also quite broad.Recently,the study of facial expression recognition technology by emotion calculation has attracted attention.Facial expression recognition tasks generally consist of four parts:obtaining facial images,image preprocessing,feature extraction,and feature classification.Facial expression images are generally obtained directly from related data sets.Therefore,the step of image preprocessing is first introduced,followed by a sequential introduction of traditional research methods and deep learning methods in feature extraction and feature classification.Finally,a comprehensive elaboration of the data sets,development trends and challenges related to facial expression recognition is described.
作者
刘建华
唐雷
LIU Jianhua;TANG Lei(China Institute of Art Science and Technology,Beijing 100012,China;Science and Technology Development Department,China Academy of Information and Communications Technology,Beijing 100191,China)
出处
《信息通信技术与政策》
2022年第8期89-96,共8页
Information and Communications Technology and Policy
关键词
人工智能
人脸表情识别
特征提取
特征分类
artificial intelligence
facial expression recognition
feature extraction
feature classification
