基于AdaBoost和LLE的视频人脸特征提取研究  

AdaBoost and LLE for face feature extraction techniques in video sequence

作　　者：王刚[1] 周激流[1] 何坤[1] 

机构地区：[1]四川大学计算机学院

出　　处：《四川大学学报（自然科学版）》2008年第3期512-516,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金(60272095);教育部博士点基金(20020610013)

摘　　要：特征提取是生物特征识别的关键环节.本文提出了一种基于AdaBoost和LLE的视频人脸特征提取方法.其大致思路是用VFW技术对视频图像进行采集、接着将AdaBoost算法对采集的图像进行人脸检测,最后应用LLE算法对检测到的人脸图像进行降维并提取出特征.项目实践表明,该方法具有便捷性、实用性和有效性.Feature extraction is a key component of biometrics. This paper puts forward a scheme based on AdaBoost and LLE from the video sequences. The basic idea for the schemes is that frame images is captured from video sequence by FVW, face detection for that images is carried out according to AdaBoost algorithm, and Dimensionality Reduction and feature extraction for Located face image is implemented by LLE. Experimental results indicate that this scheme presented in this paper is convenient, practicable and available.

关 键 词：视频序列 视频采集 ADABOOST LLE 特征提取 人脸库 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]