FACE DETECTION USING MODIFIED FDA-SVM METHOD
Hanumantha Reddy T, K. Karibasappa, Avula Damodaram · International Journal of Machine Intelligence · 2009
This paper proposes a method for face detection using Modified Fisher Discriminant Analysis (FDA) and Support Vector Machine (SVM). It is a three layer architecture system that identifies all image regions which contain face. The face detection is a preprocessing stage for an automatic face recognition system. At the first stage, the Modified Fisher Linear Discriminant Analysis (FDA) classifies the input pattern into three classes: a face class, undecided class and non-face class. At the next stage, the SVM classifies the undecided class or non face class as either face or non-face class. In the Final stage, FDA-SVM detects the face class if any sub image region falsely judged as non-face class. This system alleviates the problem of false positive rate. The experimental result shows that the proposed approach outperforms some of the existing face detection methods.