摘要
传统人脸识别技术通常在待识别人数不多时的准确率要比待识别人数多时的准确率高[1]。为了增加识别的准确性,论文提出了一种基于深度神经网络(DNN)的人脸识别方法。该方法主要涉及两方面,一是使用DNN对训练集进行特征提取;二是将提取的特征图片输入神经网络进行训练及识别。基于Python+OpenCV进行实验,通过改变权重衰减系数、卷积核数目、非线性激活函数等参数来提高卷积神经网络的性能。论文会将对卷积神经网络的参数进行大量的实验对比分析,并选取最合适的参数构建一个能够识别人脸的深度神经网络。
Traditional face recognition technology usually has a higher accuracy rate than the number of people waiting to be identified when the number of people to be recognized is. In order to increase the accuracy of recognition,this paper proposes a face recognition method based on deep neural network(DNN). This method mainly involves two aspects,one is to use DNN to extract the feature of the training set,the two is to train and identify the extracted feature pictures into the neural network. Based on Python+OpenCV,the performance of the convolution neural network is improved by changing the weight attenuation coefficient,the number of convolution kernel,nonlinear activation function and so on. In this paper,the parameters of the convolution neural network are compared and analyzed,and the most suitable parameters are selected to construct a deep neural network that can identify the face.
作者
王敏
WANG Min(Communication and Information System,Wuhan Research Institute of Posts and Telecommunic,Wuhan 430070)
出处
《计算机与数字工程》
2020年第2期433-436,466,共5页
Computer & Digital Engineering
关键词
人脸识别
特征提取
深度神经网络
face recognition
feature extraction
deep neural network