Facial Recognition Based on DWT – HOG – PCA Features with MLP Classifier

Facial Recognition System has been widely used in various applications. Nevertheless, their efficiency rate fell dramatically when they were applied under unrestrained environments like the position of face, expression or illumination change. Because of these factors, it is essential to measure and calculate the performance rate of the dissimilar feature extraction techniques robust to such transformations in order to further integrate to a Facial Recognition System. This paper studies and evaluates the Histogram of Oriented Gradients method as a feature extraction method in order to deal with the abovementioned transformations. The study consists of four main phases: face detection, preprocessing, features extraction, and classification. Preprocessing is used to enhance the images by using the techniques of digital image processing. Feature extraction is used to get features from facial images based on the concept of Histogram Oriented Gradient feature that is applied to the facial image after conversion by means of the Discrete Wavelet Transform and vector reduction with the help of the Principle Component Analysis technique. Artificial neural network with the Back Propagation algorithm is used in training and testing as a classifier of the facial images to help recognize the face. To measure the performance of the method under consideration, some experiments were implemented using two datasets: ORL containing 400 facial images of 40 individuals that achieved the accuracy rate about 99.1%, and FERET containing 912 facial images of 152 individuals, which helped achieve accuracy rate about 94.5% at the multilayer perceptron neural network classifier.