A Novel Facial Recognition Method Based On NSCT Transformation And Deep Learning Neural Network

Face recognition technology includes the field of computer vision and pattern recognition, which mainly involving computer science, pattern recognition, recognition, artificial intelligence and visual psychology theory, etc. With the rapid development of computer information technology, the face recognition technology is widely used in character recognition, information security and monitoring and other fields. Because the influence of face is affected by the factors such as light, face angle, hair occlusion and other interference factors, the face recognition technology is very complexity and difficult.For this problem,This paper has proposed a face recognition algorithm based on NSCT transform and deep learning neural network. In this algorithm, the face image will be pre processed firstly, which is including histogram equalization, image correction and image median filtering. After image preprocessing, we can get a better quality face image. And then we can get the high and low frequency characteristic data of face image through NSCT transform, which has the characteristics of multi-resolution, localization, multi direction and translational invariance. At last, the algorithm will get the lower dimension by the PCA dimensionality reduction method. Finally, through the characteristic data training and learning of depth learning neural network. The neural network is composed of two optimal learning efficiency RBM and a GRNN neural network, which is a new neural network to train and study the human face images by the new deep learning neural network.At last, the algorithm is simulated by the MATLAB simulation platform, the performance of the proposed algorithm is verified by using ORL and YALE face image database. Compared with other face recognition algorithms, the proposed algorithm not only has a high face recognition rate, but also takes up a smaller feature vector storage space. By the simulation, the algorithm proposed in this paper with face recognition rate is 97%, in the presence of external disturbances, the face recognition rate is still more than 89%, with good performance in face recognition.