Dynamic Face Recognition System Using Histogram of Oriented Gradients and Deep Neural Network

Abstract

AbstractFace recognition is a rapidly growing field. Face recognition is the ability to recognize and detect a person based on their visual features. Many methods are already proposed in which the face recognition system can recognize only a fixed set of people. This paper provides a revolutionary face recognition system implementation process that can recognize a large number of persons whose facial features may be derived from an input image. Face detection and recognition are the two phases of the human face recognition technique. In this model, face recognition is achieved by using histogram of oriented gradients and deep neural network. Histogram of oriented gradients (HOG) extracts all the important features from an image. HOG coupled with support vector machine (SVM) classifier detects the face from an input image. The localized face image is passed to the regression tree for facial landmark detection to obtain a centered face image which is fed as input to the deep neural network for face recognition. The proposed model gives an accuracy of 94.44% and is also implemented on video. This model can be used for biometric attendance systems.KeywordsHistogram of oriented gradientsFace detectionConvolutional neural networkDeep learningLinearSVM