Human Face Recognition Using Discriminant Analysis

Abstract

In the present research, a face recognition method is proposed based on the concept of linear discriminant analysis (LDA) method. The LDA requires input some of image models to analyze and discriminate them, the newly proposed idea is the use of a number of textural features instead of face image pixels to be input the LDA procedure. The employed textural features were ten, which are computed for each face image using the grey level co-occurrence matrix (GLCM) method. The proposed face recognition method consists of two phases: enrollment and recognition. The enrollment phase is responsible for collecting the features of each face image to be a comparable models stored in the database, while the recognition phase is responsible on comparing the extracted features of input unknown face with that stored in the database. The comparison results a number of percentage values, each refers to the similarity between the input unknown face with the models in the database. The recognition decision is then issued according to the comparison results. The results showed that the system performed the recognition test with a recognition percent of about 94%, whereas the validation test showed that the system performance was about 92%. Frequent practices showed that the behavior of the recognition is acceptable and it is enjoying with the ability to be improved.