基于孪生对齐卷积神经网络的面部表情识别
Facial Expression Recognition Based on Siamese Aligned Convolutional Neural Network
杨晓峰1
作者信息
- 1. 山西工程科技职业大学计算机工程学院,山西 晋中 030600
- 折叠
摘要
分析面部表情在医疗、教育、刑侦、交通以及人机交互等方面有很高的实用价值.设计了孪生对齐卷积神经网络模型,将面部图像左右分割后进行表情识别,达到了较为理想的识别准确率.FER2013 和CK+数据集上的测试结果表明,准确率分别提高了 0.9 和 0.6 个百分点.实验结果说明,该模型在识别面部表情时是有效的.
Abstract
Facial expressions can better reproduce human real thoughts.Facial expressions analysis has great practical value in healthcare,education,criminal investigation,transportation and human-computer interaction.In this paper,the Siamese aligned neural network is designed to achieve a more satisfactory recognition accuracy by segmenting facial im-ages left and right.On FER2013 and CK+,the accuracy improved by 0.9 and 0.6 percentages,respectively.The experi-mental results illustrate that our model is effective in recognizing facial expressions.
关键词
表情识别/孪生神经网络/特征对齐/分割Key words
facial expression recognition/Siamese neural network/feature alignment/segmentation引用本文复制引用
基金项目
山西工程科技职业大学科研项目(KJ202201)
出版年
2024
NETL
NSTL
万方数据