基于SVM特征层融合的集成性身份识别模型  

A FEATURE-LEVEL SVM FUSION BASED INTEGRATION IDENTIFICATION MODEL

作　　者：周伟芳[1] 叶学义[1] 何文韬[1] 刘一锐[1] 

机构地区：[1]杭州电子科技大学通信工程学院浙江杭州310018

出　　处：《计算机应用与软件》2013年第9期81-83,106,共4页Computer Applications and Software

基　　金：国家自然科学青年基金项目(60802047,61001216);浙江省自然科学基金项目(R1090138);浙江省科技计划重点项目(2008C21092)

摘　　要：多生物特征融合识别是提高身份识别性能的有效方法,提出一种基于特征层融合的识别模型。以人脸与虹膜为例,首先对这两种图像分别进行预处理,然后对人脸采用PCA、Fisherface方法,对虹膜采用小波变换、多通道Gabor滤波方法,进行特征提取得到各自的初始特征向量,由此生成融合特征向量,最后利用SVM对融合特征向量进行分类,实现特征层融合识别。仿真实验结果表明,同等条件下,该融合方法的识别效果优于常见识别方法。Multi-biometric fusion recognition is an effective method to improve the identification performance, In this paper, we propose a kind of feature-level fusion based recognition model. Taking the face and iris for example, firstly it pre-processes these two kinds of image respectively, then applies PCA and Fisherface on face, and wavelet transform and multi-channel Gabor filter on iris to get their initial feature vectors respectively from feature extractions, thus generates the fusion feature vector, finally it classifies the fusion feature yector with SVM classification, and achieves feature-level fusion recognition. Simulation experimental results show that, under the same condition, this fusion method performs better in recognition effect than the common method.

关 键 词：SVM 特征层融合 生物特征识别 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]