Face detection based on dimension reduction using probabilistic neural network and Genetic Algorithm

Past work on face detection has emphasized the issues of feature extraction and classification, however, less attention has been given on the critical issue of feature selection. We consider the problem of face and non-face classification from frontal facial images using feature selection and neural networks. We argue that feature selection is an important issue in face and non-face classification. Automatic feature subset selection distinguishes the proposed method from previous face classification approaches. First, principal component analysis (PCA) is used to represent each image as a feature vector (i.e., eigen-features) in a low-dimensional space, spanned by the eigenvectors of the covariance matrix of the training images (i.e., coefficients of the linear expansion).Then we consider linear discrimination analysis (LDA) to achieve a comparison result between these two methods of dimension reduction. Genetic algorithm (GA) is then used to select a subset of features from the low-dimensional representation by removing certain eigenvectors that do not seem to encode important information about face. Finally, a probabilistic neural network (PNN) is trained to perform face classification using the selected eigen-feature subset. Experimental results demonstrate a significant improvement in error rate reduction.