摘要
针对(2D)2PCA(two-dimensional principal component analysis)提取的特征脸精度变低的问题,本文引入插值法在特征向量之间插入新的向量,以提高特征信息的显示度;针对传统的神经网络存在学习效率低、收敛速度慢和容易陷入局部极小值的问题,本文使用一种基于权值缓慢变化的粒子群算法(particle swarm optimization with slowly changing weights,WSCPSO)优化神经网络权值.实验表明:两种算法的结合能够大大地提高识别率.
To solve the problem of the low accuracy of face features extracted by two-dimensional principal component analysis((2 D)2PCA),we introduce an interpolation method for inserting new vectors between feature vectors to improve the display of feature information.To optimize the weights of neural networks,we use a particle swarm optimization algorithm with slowly changing weights.Experimental results show that the combination of these two algorithms can greatly improve the recognition rate.
作者
牛冰川
文成林
NIU Bingchuan;WEN Chenglin(School of Automation,Hangzhou Dianzi University,Hangzhou 310018,China)
出处
《信息与控制》
CSCD
北大核心
2021年第3期350-355,365,共7页
Information and Control
基金
国家自然科学基金资助项目(61751304,61673318,U1664264)
浙江省自然科学基金资助项目(LZ16F03000)。
关键词
神经网络
特征提取
优化算法
特征投影框架
人脸识别
neural network
feature extraction
optimization algorithm
feature projection framework
face recognition