摘要
由于人脸面貌特征与性别存在着一定的不确定性,提出了基于模糊隶属度的人脸图像性别识别。用对光照、灰度变化具有较强鲁棒性的局部二进制模式(LocalBinary Pattern,LBP)提取人脸特征,首先将人脸均分为多个子窗口,对所有子窗口提取LBP直方图,然后将这些直方图顺次连接来描述人脸。细致推导了适用于人脸图像性别识别的模糊函数,根据最大隶属度原则,来识别人脸的性别。在FG-NET人脸库及自建的FID人脸库中进行了实验,取得了96%的最高识别率。
Because of the greater uncertainty exists in both face features and gender, a novel method based on fuzzy membership degrees for gender recognition of face image is proposed. Face features are extracted with Local Binary Pattern which is robust to the il- lumination change and gray variations. First, face image was divided into many of the same size windows, features are extracted with LBP method for each window, second, features of all windows are joined up in turn aim is to describe the face. The fuzzy function is appropriate for gender recognition of face image was derived rigorous. The principle of maximum membership degree is used to gender recognition, the experiments were conducted in FG-NET face database and own FID face database, the highest recognition rate of 96% is achieved.
出处
《软件》
2012年第8期28-31,50,共5页
Software
基金
山西大同大学科研项目(2012K8)~~
关键词
局部二进制模式
特征提取
模糊隶属度
性别识别
local binary pattern features extraction fuzzy membership degrees gender recognition
