摘要
针对传统KNN算法计算量大、识别率低的问题,提出一种加权K最近邻法(KNN)结合随机森林(RF)的表情识别方法。首先通过监督下降方法(SDM)提取人脸特征点,然后计算样本间的平均距离,并借此划分测试样本,结合加权KNN与随机森林的特点,对不同样本采用不同的分类器,最后采用JAFFE表情数据库进行实验。结果表明,改进后的方法不仅识别率更高,而且简化了计算复杂度。
To solve the problem of the large computation and low recognition rate which caused by the K-Nearest Neighbor(KNN)algorithm,a new approach for facial expression recognition based on weighted K-Nearest Neighbor and Random Forest(RF)is presented in this paper.First of all,the features of the static facial expression image are extracted by the supervised descent method(SDM),then the average distance between samples are calculated and used to divide test samples,different classifiers for different test samples are adopted based on the characteristic of weighted KNN and RF.Finally,the results of experiment on JAFFE database show that the proposed algorithm can not only achieve better recognition rate,but also simplify the computation complexity.
作者
冯开平
赖思渊
FENG Kai-ping;LAI Si-yuan(School of Computing,Guangdong University of Technology,Guangzhou 510006,China)
出处
《软件导刊》
2018年第11期30-33,共4页
Software Guide
基金
广东省自然科学基金项目(2015A030310112)
广东省科技计划项目(2016A040403110)
关键词
表情识别
K最近邻
随机森林
expression recognition
K-Nearest Neighbor
random forest
