The recognition of human faces by computer

Abstract

Included are two main classes, i.e. frontal face and profile in face recognition. This paper describes the former class and represents a method of feature selection based on psychological test conclusions and K-L transformation to remove the correlativity of the original features, and a new discriminant criteria called Distance Fuzzy Diagnosis Maximum 1-NN, which solves the problem of error decision at the second or third place of order-ranked distance list in Minimum Distance Criteria. Through the identification of individual faces of seventy pictures and ten subject's facial image with different facial orientation, the correct accuracy achieves 97.5% and 90% respectively.<>