基于Self-adjust网络模型的人脸图像情感分析方法  

A face image sentiment analysis method based on Self-adjust network model

作　　者：邓亚萍 王新[1] 尹甜甜 王婷 郑承宇 DENG Yaping;WANG Xin;YIN Tiantian;WANG Ting;ZHENG Chengyu(School of Mathematics and Computer Science,Yunnan Minzu University,Kunming 650500,China)

机构地区：[1]云南民族大学数学与计算机科学学院,云南昆明650500

出　　处：《河南工程学院学报（自然科学版）》2022年第1期61-65,75,共6页Journal of Henan University of Engineering：Natural Science Edition

基　　金：国家自然科学基金(61363022);云南民族大学研究生创新基金项目(SJXY-2021-003)。

摘　　要：人脸表情具有丰富的情感内涵,是情感分析的一个重要研究方向。模糊的面部表情及标注者的主观性所带来的不确定性,给情感分析研究带来了挑战。鉴于此,提出了一种基于Self-adjust网络模型的人脸图像情感分析方法。首先用人脸对齐方法进行图像预处理,然后利用注意力机制来处理Focal损失加权,再对其进行秩正则化排序,最后通过重新分类对有误标签进行矫正,并用实验验证了该方法的有效性与优越性。该方法在准确率这个评价指标上有所提高,能够有效抑制人脸图像情感分析的不确定性,防止深层网络对不确定的人脸图像进行过拟合。Facial expression has rich emotional connotations and is an important research direction of sentiment analysis.Due to the uncertainty brought by fuzzy facial expressions and the subjectivity of the annotator,the sentiment analysis research has been challenged.In view of this,this paper proposes a face image sentiment analysis method based on the Self-adjust network model.First,face alignment was used for image preprocessing,then attention mechanism was used to process Focal loss weighting,then rank regularization sorting was performed on it,and finally,mislabeled images were corrected by reclassification,the experimental results verify the effectiveness and superiority of this method.This method has been improved in the accuracy of evaluation indexes,which can effectively suppress the uncertainty of face image sentiment analysis and prevent the deep network from over-fitting uncertain face images.

关 键 词：Self-adjust网络模型 人脸对齐 注意力机制 Focal损失加权 情感分析 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]