Research Of Face Recognition Based On Convolutional Neural Network

Face recognition has always been a hot spot in the field of pattern recognition research because it is a biometric recognition technology which is not easy to copy.However,there are many restrictions in the traditional face recognition technology,so it is not easy widely used.In recent years,with the continuous development of deep learning technology,convolutional neural networks with its excellent feature extraction ability has been widely used in pattern recognition research.Face recognition based on convolutional neural networks has many advantages compared with traditional face recognition.First of all,the traditional face recognition is based on the methods which is need extract features depending on the subjectivity of researchers.Secondly,the convolution neural network can extract the hiden features and the expression of the characteristics is stronger.In this paper,a Full-convolutional layer neural network structure has been designed that can be applied to the research of face detection.In the research of face recognition based on convolutional neural networks,spatial pyramid pool(SPP)layer is introduced between the full-connection layer and the last convolutional layer instead of the traditional pool layer.SPP layer was proposed to ensure arbitrary input image size and turn the feature maps into a fixed length vector.The results show that the network based on the full-convolutional layers is not only process image at a high speed,but also it is feasible.And that lack of the negative samples will lead to an increase in false alarm rate.The convolutional neural network based on spatial pyramid pooling can effectively extract and express hiden facial features in recognition,achieve higher recognition rate.The research also shows that network have an excellent performance can be determined by the best type of activation function and the range of relevant parameters.