Smart Approach Based on CNN and Shearlet Transform for Age Prediction

Abstract

AbstractThe age and gender, which are basic features of the face, play a very fundamental role in social interactions, which makes age and gender estimation from a single photo of the face an important process in smart applications. In our last research work, we proposed an approach based on convolutional neural network (CNN) combined with shearlet transform (ST), in order to design a gender recognition system. In this work, we will exploit this approach, by proposing an age recognition system. Based on a CNN for classification and a shearlet transformation to extract the essential features of the face, our approach guarantees a better recognition rate compared to using a simple CNN. This improvement in the recognition rate is the result of extracting multiple frequency views from a face photo which increases the input sample in the CNN learning phase.KeywordsAge predictionShearlet transformCNNClassificationRecognitionFeatures extraction