即時人臉偵測與辨識

Real Time Face Detection and Recognition

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為對象，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為人臉辨識之決策法則。

This paper presents a real-time face detection and recognition systemusing the database of AT&T Laboratories Cambridge, Georgia Institute ofTechnology, California Institute of Technology and a database created bythe authors. In addition, the effect of training samples on face recognitionis investigated in this paper. The face detection algorithm was implementedin a Matlab/Simulink environment to reduce the running time todetect 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, thenearest feature line method and linear discriminant analysis.