多尺度注意力机制下的人脸表情识别算法设计
Design of face expression recognition algorithm under the multi-scale attention mechanism
蒋文豪1
作者信息
- 1. 重庆航天职业技术学院智能信息工程学院,重庆 400021
- 折叠
摘要
为了优化在人脸表情较模糊情况下的识别效果,并更好地获取表情的表征数据,设计一种多尺度注意力机制下的人脸表情识别方法.对人脸表情图像进行缩放与扩充预处理操作,从图像中提取人脸表情解耦表征皮沟数据,通过卷积神经网络对提取到的解耦表征皮沟数据进行特征捕捉.引入了多尺度注意力机制,有选择性地关注重要的表情特征.同时,利用多通道的表情识别方法,自适应地提取人脸组件区域内的表情信息,从而识别人脸表情.实验分析结果表明,所提方法在四类不同表情标签对应的人脸表情识别召回率始终高于对照组,均达到了98%以上,识别效果优势显著.
Abstract
In order to optimize the recognition effect of blurred face expression and better obtain the representation data of fa-cial expression,a face expression recognition method under the multi-scale attention mechanism is designed.By zoom and expan-sion,the face expression decoupling representation feature data is extracted from the image,and the extracted decoupling represen-tation feature data is captured through the convolutional neural network.Introduce multiscale attention mechanisms to selectively focus on important expression features.at the same time.Using the multi-channel expression recognition method,the expression in-formation in the face component area is adaptively extracted to identify facial expressions.The experimental analysis results show that the recall rate of face expression recognition corresponding to the four different types of expression labels is always higher than that of the control group,which is more than 98%,and the recognition effect has significant advantages.
关键词
多尺度注意力/人脸特征/表情识别/皮沟特征/神经网络Key words
multi-scale attention/face features/expression recognition/skin features/neural network引用本文复制引用
基金项目
重庆市教育委员会科学技术研究项目(KJZD-K202003001)
出版年
2024
NETL
NSTL
维普
万方数据