A sparse representation face classification method based on convolutional neural network

Abstract

Convolutional neural network has a very important role in feature extraction, with stronger feature learning and feature expression ability. Therefore, to address the lack of robustness of sparse representation and the changes of environment and age in the face feature classification, at the feature level, this paper studies the convolutional neural network. We proposed a face feature extraction method based on convolutional neural network and then face classification using sparse representation. This method combines convolutional neural network extraction and sparse representation, taking full advantages of convolutional neural network in deep feature extraction and sparse representation in face feature recognition classification. The experimental results show that acquiring features through the convolutional neural network can represent faces linearly, which is more robust, and satisfies the assumption conditions of the face classification method based on sparse representation. The final experimental results show that the proposed face classification method is 9%~20% higher than the traditional feature extraction method.