Mohammad
Hossein Chamani
Abstract
Face recognition is used as one of the most successful biometric methods due to the availability of advanced resources such as faster processors and higher memory and providing intelligent methods based on the power of these resources. Nevertheless, there are still many challenges in this area. The face plays an important role in the transmission of emotions and carries the characteristics hidden in it, the identity of individuals. Face recognition has been added to some control devices, security, welfare, criminal identification, and many other areas, which is the main motivation for research in this field. In this paper, the DCSFR method is presented to pay attention to the main features of the face such as eyes, lips, mouth, and nose, which is the main novelty of this work, to get higher accuracy or speed than the previously existing methods. In this approach, instead of using general information in face recognition, facial components such as eyes, nose, mouth are separated into another image, and face classification operations (deep learning by convolution neural network) are performed on separated components. The results show that the computational cost with the proposed method is reduced by about 70%. Also, it can be achieved that CNN does not perform as well as the complete picture of the disassembled components.
Keywords
Deep Learning; Convolutional Neural Network; Machine Vision; Facial Components Segmentation; Facial Recognition.
