即時人臉偵測與辨識

Time Face Detection and Recognition

李建興(Chien-Hsing Lee) ； 林應璞(Ying-Pu Lin) ； 游凱倫(Kai-Lun You)

《技術學刊》 24卷2期 (2009/06) Pp. 131-141

https://doi.org/10.29507/JT.200906.0005

摘要

本文以AT&T Laboratories Cambridge, Georgia Institute of Technology, California Institute of Technology與作者自建人臉影像資料庫爲對象，進行即時人臉偵測與辨識，並討訓練樣本數對人臉辨識之影響。首先，以Matlab/Simulink進行即時人臉偵測，用以降低人臉偵測所需時間，再使用Haar小波、主成份分析法與改良式主成份分析法擷取人臉特徵，最後以歐氏距離決策法、最接近特徵線決策法與線性鑑別式分析法作爲人臉辨識之決策法則。

關鍵字

主成份分析法 ； 改良式主成份分析法 ； 線性鑑別式分析 ； 最接近特徵線法

並列摘要

This paper presents a real-time face detection and recognition system using the database of AT&T Laboratories Cambridge, Georgia Institute of Technology, California Institute of Technology and a database created by the authors. In addition, the effect of training samples on face recognition is investigated in this paper. The face detection algorithm was implemented in a Matlab/Simulink environment to reduce the running time to detect a face. Then, the facial feature extraction part uses Haar wavelets, principal component analysis and improved principal component analysis. Finally, the facial recognition rule is based on Euclidean distance, the nearest feature line method and linear discriminant analysis.

並列關鍵字

principal component analysis ； improved principal component analysis ； linear discriminant analysis ； nearest feature line method