Hybrid Multi-class Learning and Augmentation Techniques for Human Age Determination

Age identification plays a critical role in daily life. It has many applications, such as unmanned stores, monitoring, marketing, and autonomous vehicles. A challenge of age estimation is that people have little change in the face during the middle age. Therefore, in this work, we propose a sophisticated age estimation architecture that adopts many advanced techniques to improve the accuracy. In order to address the problem that the face changes slightly in middle age, we classify the age into 7 classes and apply different classification models for different classes. The proposed system is divided into three stages. The first stage is to perform augmentation, including feature map construction, face detection, and cropping. The second stage is to classify the inputs into seven classes of ages through a machine learning technique that is a combination of deep learning and the support vector machine (SVM). The third stage is to train seven Deep Neural Networks (DNN) models for the seven classes and then predict the age explicitly. The input for each DNN model contains the probability of each age group estimated by the previous stage. Simulations show that the proposed algorithm outperforms state-of-the-art methods and has accurate results in age estimation.