摘要
针对人脸识别问题,提出一种基于多自由度神经元模型的人脸识别算法.首先,给出多自由度神经元模型及其神经网络的构造算法;然后,将算法用于UMIST人脸库和YALE人脸库,同时在计算中给出了一种简单有效的减少光照影响的预处理的方法;最后将本文算法与SVM算法进行比较,证明了算法的有效性.
In this paper, a novel algorithm based Multi-degree of Freedom neuron is proposed to deal with face recognition. Firstly, the Multi-degree of Freedom neuron model and its neural network conformation algorithm are given; Secondly, the algorithm is ap- plied to the UMIST and Yale face database. A new pretreatment method is proposed to reduce the effect of light at the same time; lastly, the algorithm is compared to the SVM and its effectiveness is proved by the experiments.
出处
《小型微型计算机系统》
CSCD
北大核心
2012年第12期2693-2695,共3页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60871093)资助
浙江省教育厅科研项目(Y200907801)资助
关键词
人脸识别
多维空间
仿生信息学
多自由度
神经网络
face recognition
multi-dimensional space
biomimetic informatics
multi-degree of freedom
neural networks