基于蚁群智能和支持向量机的人脸性别分类方法  

Gender Classification Based on Ant Colony and SVM for Frontal Facial Images

作　　者：燕忠[1] 袁春伟[1] 

机构地区：[1]东南大学生物医学工程系分子与生物分子电子学教育部重点实验室,南京210096

出　　处：《电子与信息学报》2004年第8期1177-1182,共6页Journal of Electronics & Information Technology

基　　金：国家自然科学基金重点项目(No.69831010)资助

摘　　要：蚁群优化算法是根据自然界中蚂蚁能够将食物以最短路径搬回蚁巢这一智能行为而提出的一种新颖的进化算法,该算法不仅具有很好的鲁棒性,良好的正反馈特性,而且具有并行分布计算的特点.同时,支持向量机又是一种基于结构风险最小化原理的机器学习技术,具有很强的学习泛化能力,为此,文章提出了基于蚁群优化算法和支持向量机的人脸性别分类的方法.首先,通过KL变换降低人脸性别特征的维数,并根据特征值按照从大到小的顺序进行排列,然后采用10-交叉确认技术,用蚁群优化算法对人脸性别特征面进行选择,以对支持向量机进行学习、训练和测试.实验表明,与其他分类算法相比较,这种方法不仅图像处理简单,实用性强,而且正确识别率特别高.Ant Colony Optimization (ACO) is a novel evolutionary algorithm derived from the foraging behavior of real ants of nature, which can find the shortest path between a food source and their nest. The main characteristics of ACO are robustness, positive feedback and distributed computation. And at the same time, Support Vector Machine (SVM), based on structure risk minimization principle, has the better performance and the better generalization ability. According to these, a gender classification using SVM and ACO is presented. Firstly, to reduce the dimensionality of the face images, the principal component coefficients of all images are calculated through Karhunen Loeve transform. Then, the eigenvectors are sorted in the descending order of eigenvalues. Secondly, ACO decides which eigenvectors will be used. After ACO's feature selection, the SVMs are trained and tested for gender classification. Deserving the best optimal features with highest accuracy rate, the next validation is continued until 10-fold cross-validations are completed. The experiments indicate that the proposed gender classification system based on ACO and SVM is more practical and efficient in comparison with others.

关 键 词：性别分类 支持向量机 蚁群优化算法 KL变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]