判别性完全局部二值模式人脸表情识别

发布时间：2018-03-21 20:47

本文选题：完全局部二值模式(CLBP)　切入点：有判别力的完全局部二值模式(disCLBP)　出处：《计算机工程与应用》2017年04期 　论文类型：期刊论文

【摘要】：针对完全局部二值模式(CLBP)存在直方图维数过高和特征冗余,会导致识别速度降低和识别率低的问题,提出基于有判别力的完全局部二值模式(Discriminative completed LBP,dis CLBP)的人脸表情识别算法。首先,对人脸表情图像进行预处理获得表情子区域;然后提取表情子区域和整幅图像的dis CLBP特征,针对不同的表情筛选出不同的表情特征,再将筛选出的表情子区域特征直方图融合;最后用最近邻分类器进行分类识别。该算法在CK人脸表情库上进行实验的平均识别率为97.3%。
[Abstract]:In view of the problem of high histogram dimension and feature redundancy, the problem of low recognition speed and low recognition rate can be caused by the completely local binary mode CLBP. A face expression recognition algorithm based on discriminative completed discriminative completed is proposed. Firstly, the facial expression sub-region is obtained by preprocessing the facial expression image, and then the dis CLBP feature of the expression sub-region and the whole image is extracted. According to different facial expressions, different facial features are selected, and then the histogram of the selected facial expression sub-regions is fused. Finally, the nearest neighbor classifier is used to classify and recognize. The average recognition rate of the algorithm in the CK facial expression database is 97.3%.
【作者单位】：江南大学物联网工程学院智能系统与网络计算研究所;
【基金】：国家自然科学基金(No.61202312,No.61170121) 教育部留学回国人员科研启动基金项目
【分类号】：TP391.41

【相似文献】