计算机科学 ›› 2015, Vol. 42 ›› Issue (12): 302-306.
范燕,吴小俊,邵长斌,宋晓宁
FAN Yan, WU Xiao-jun, SHAO Chang-bin and SONG Xiao-ning
摘要: 针对传统特征提取方法和BP神经网络相结合而存在的缺点,提出一种新的分类器模型“PCABP网”。首先利用PCA特征向量来初始化PCABP网的初始层权值矩阵,由此新模型的初始层起到取代PCA进行特征提取的作用。其次在训练过程中通过GHA和GD算法对初始层投影权值矩阵进行动态调节来优化特征向量。该方法从源头样本来优化“类别分离”和“特征提取”,找到对样本降维和分类的最佳契合点,以此来替代传统模式识别中“首先单独特征提取,其次利用分类器分类”的方式。在FERET人脸库上的实验结果验证了该方法的有效性。
[1] 边肇祺,张学工.模式识别(第二版)[M].北京:清华大学出版社,1999 Bian Zhao-qi,Zhang Xue-gong.Pattern Recognition (Second Edition) [M].Beijing:Tsinghua University Press,1999 [2] 王成,郭飞,赖雄鸣,等.典型代数统计的人脸特征提取融合[J].小型微型计算机系统,2014,5(7):1662-1666 Wang Cheng,Guo Fei,Lai Xiong-ming,et al.Face Feature Extraction Fusion Based on Classical Algebraic Statistics[J].Mini-Micro Systems,2014,5(7):1662-1666 [3] 方晶晶,李振波,姜宇.人体肤色区域的自适应模型分割方法[J].计算机辅助设计与图形学学报,2013,5(2):229-234 Fang Jing-jing,Li Zhen-bo,Jiang Yu.Human Skin Color Region Segmentation Based on Adaptive Model[J].Journal of Compu-ter-Aided Design & Compu-ter Graphics,2013,5(2):229-234 [4] Hinton G E,Salakhutdinov R R.Reducing the dimensionality of data with neural networks.[J].Science(New York,N.Y.),2006,313(5786):504-507 [5] 余凯,贾磊,陈雨强,等.深度学习的昨天、今天和明天[J].计算机研究与发展,2013,50(9):1799-1804 Yu Kai,Jia Lei,Chen Yu-qiang,et al.Deep Learning:Yesterday,Today,and Tomorrow[J].Journal of Computer Research and Development,2013,50(9):1799-1804 [6] 聂仁灿,姚绍文,周冬明.基于简化脉冲耦合神经网络的人脸识别[J].计算机科学,2014,1(2):297-301 Nie Ren-can,Yao Shao-wen,Zhou Dong-ming.Face Recognition Using Simplified Pulse Coupled Neural Network[J].Computer Science,2014,1(2):297-301. [7] 刘贵松,王晓彬.采用自适应GHA神经网络的分类器设计[J].电子科技大学学报,2007,6(6):1241-1244 Liu Gui-song,Wang Xiao-bin.Classifier Design Using Adaptive GHA Neural Networks[J].Journal of University of Electornic Science and Technology of China,2007,6(6):1241-1244 [8] 杨存祥,朱堔,解豪杰.基于RPROP神经网络算法的异步电动机故障诊断[J].电力自动化设备,2012,2(1):80-83 Yang Cun-xiang,Zhu Chen,Xie Hao-jie.Fault Diagnosis Based on RPROP Neural Network for Asynchronous Motor[J].Electric Power Automation Equipment,2012,2(1):80-83 [9] Liu X,Chen T,Kvijava Kumar B V.Face authentication formultiple subjects using eigenflow [J].Pattern Recognition,2003,36(2):313-328 [10] Sander T D.Optimal unsupervised learning in a single-layer li-near feed forward neural network [J].Neural Networks,1989,2:459-473 [11] Alex L,Sandor C,Der Nasser Z,et al.Dual-band passive infrared imagery for automatic clutter rejection [M].ICIP,2000 [12] Belhumeur P N,Hespanha J P,Kriengman D J.Eigenfaces vs.Fisherfaces:Recognition Using Class Specific Linear Projection [J].IEEE Trans.Pattern Analysis and Machine Intelligence,1997,9(7):711-720 |
No related articles found! |
|