结合主方向和SVM的人脸表情识别

发布时间：2018-02-27 19:21

本文关键词： 人脸表情识别 LBP 支持向量机(SVM) 主方向　出处：《廊坊师范学院学报(自然科学版)》2016年04期 　论文类型：期刊论文

【摘要】：随着人工智能和人脸表情识别技术的不断发展,现有的人脸表情识别方法取得了不错的效果。但是,对于存在姿态变化较大的表情库,如何提取稳健有效的表情特征,仍然是人脸表情识别需要面对的问题。受尺度不变变换特征的启发,在块LBP特征提取中引入主方向信息,使得提取的特征具有一定的旋转不变性,利用投票决策SVM分类器最终完成了表情识别任务。通过在公共表情库JAFFE和Cohn-Kanade上进行实验,平均识别率均达到95.5%以上,且时间性能得以有效提升。
[Abstract]:With the development of artificial intelligence and facial expression recognition technology, the existing facial expression recognition methods have achieved good results. It is still a problem to face in facial expression recognition. Inspired by the invariant transformation of scale, the main direction information is introduced into the feature extraction of block LBP, which makes the extracted feature have a certain rotation invariance. Finally, the SVM classifier of voting decision is used to complete the task of facial expression recognition, and the average recognition rate is over 95.5% through the experiments on JAFFE and Cohn-Kanade, and the performance of time is improved effectively.
【作者单位】：安徽理工学校;
【分类号】：TP391.41

【参考文献】