Development of a local binary pattern descriptor for texture analysis using deep convolutional neural network

The machine learning is an important branch of artificial intelligence. In machine learning, the facial expression recognition continued a difficult and interesting topic. The majority of extant techniques are based on traditional features descriptors such as local binary patterns and its extended variants. This research paper expresses the strength of deep learning techniques that is the deep convolutional neural networks for classification of faces in selected dataset. We applied proposed Deep CNN architecture with local binary patterns and histogram of oriented gradient method. Viola Jones algorithm is applied for detection of faces from Face Recognition dataset. The features of faces are extracted by Binary Phase Component Analysis and information correlation factor gain. Then the classification of images is performed by our proposed deep convolutional neural network (Deep CNN). The performance of the model is evaluated by accuracy, precision, recall, f1-score and confusion matrix. The architecture of CNN constitutes convolutional layer, max-pooling, dense and flatten layers with dropout. The proposed architecture is validated on the Face Recognition dataset. We obtained 0.98 as accuracy which is the very high accuracy of deep CNN model for classification.