基于小波变换与脉冲耦合神经网络的人脸识别  

Face Recognition Based on Wavelet Transform and Pulse Coupled Neural Network

作　　者：何世强 刘金清 刘引 蔡淑宽 陈存弟 周晓童 邓淑敏 吴庆祥 HE Shi-Qiang LIU Jin-Qing LIU Yin CAI Shu-Kuan CHEN Cun-Di ZHOU Xiao-Tong DENG Shu-Min WU Qing-Xiang(Key Aboratory of OptoElectronic Science and Technology for Medicine of Ministry of Education, College of Photonic and Electronic Engineering, Fujian Normal University, Fuzhou 350007, Chin)

机构地区：[1]福建师范大学光电与信息工程学院医学光电科学与技术教育部重点实验室,福州350007

出　　处：《计算机系统应用》2017年第10期231-235,共5页Computer Systems & Applications

基　　金：国家自然科学基金(61179011);福建教育厅项目(JAS151254);福建师大项目(I201502019)

摘　　要：脉冲耦合神经网络(Pulse Coupled Neural Network,PCNN)是基于生物视觉特性而提出的新一代人工神经网络,它在数字图像处理及人工智能等领域具有广泛应用前景.本文通过研究PCNN理论模型及其工作特性的基础上提出了一种提取人脸特征的方法.首先利用小波变换提取人脸图像低频特征,降低人脸图像的维度,然后利用简化的PCNN提取小波低频系数重构后的人脸图像的相应时间序列,并以此作为人脸识别的特征序列.最后利用时间序列和欧式距离完成人脸的识别过程.本文通过ORL人脸库进行实验证明了该方法的有效性.Pulse Coupled Neural Network（PCNN） is a new generation artificial neural network（ANN） based on biological vision. It has wide application prospects in the field of digital image processing and artificial intelligence. In this paper, we propose a method to extract face features by studying PCNN theoretical model and its working characteristics. Firstly, the low frequency feature of face image is extracted by wavelet transform. Then, the simplified PCNN is used to extract the corresponding time series of face image reconstructed by wavelet low-frequency coefficient,which is used as the feature sequence of face recognition. Finally, the face recognition process is completed with time series and Euclidean distance. In this paper, we demonstrate the effectiveness of the method with ORL face database.

关 键 词：脉冲神经网络 人脸识别 小波变换 时间序列 欧氏距离 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]