摘要
针对支持向量机(SVM)在超平面附近进行性别识别的不准确性,引入进行加权的K近邻(KNN)算法。提出了结合加权KNN和SVM的改进KNN-SVM算法,该算法用少量已知性别样本自动确定加权KNN与SVM的最优分类阈值,并计算待识别样本和支持向量机所确定的超平面的距离,通过距离与阈值的比较进行性别识别。基于FERET人脸库进行性别实验,实验结果表明,该算法比SVM算法和不进行加权处理的KNN-SVM算法的识别率更高。
Improved KNN-SVM that combined Support Vector Machine(SVM) with weighted K Nearest Neighbor(KNN) is pre-sented to improve the accuracy of gender recognition nearby SVM hyperplane.The algorithm gets optimal threshold by a few of known gender samples,then computes the distances from the test samples to the optimal superplane of SVM in feature space,recognizes gender after comparing the distance to threshold.The experiments show that the mixed algorithm can improve the ac-curacy compared to SVM and KNN-SVM without weight value.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第4期177-179,224,共4页
Computer Engineering and Applications
基金
国家自然科学基金No.606731190
江苏大学高级专业人才科研启动基金资助项目No.05FDG020~~
关键词
人脸性别识别
支持向量机
K近邻距离分类器
最优阈值
facial gender recognition
Support Vector Machine(SVM)
K-Nearest Neighbor(sKNN)classification
optimal threshold