Classification in data mining for face images using neuro: genetic approaches

Abstract

This paper describes a method of hybrid classifier/recogniser based on Neuro-Genetic processing of face images. The use of Data Mining techniques has a legitimate and enabling ways to explore large image collections using the Neuro-Genetic approaches. Much research in human face recognition involves fronto-parallel face images, which are operated under strict imaging conditions such as controlled illumination and limited facial expressions. A novel Symmetric-Based Algorithm is proposed for face detection in still grey-level images, which acts as a selective attentional mechanism. A fusion of three face classifiers, Linear Discriminant Analysis (LDA), Line-Based Algorithm (LBA) and Kernel Direct Discriminant Analysis (KDDA), is proposed with Genetic Algorithm, which optimises the weights of neural network. It helps to extract only the essential features that effectively and successively improve the classification accuracy. The BioID face database, from BioID Laboratory, Texas, USA, has 1024 images for 22 subjects are used for analysis.