详细信息

基于栈式自编码器的FTIR光谱识别    

IDENTIFICATION OF FOURIER TRANSFORM INFRARED SPECTROSCOPY BASED ON STACKED AUTO-ENCODER

文献类型:期刊文献

中文题名:基于栈式自编码器的FTIR光谱识别

英文题名:IDENTIFICATION OF FOURIER TRANSFORM INFRARED SPECTROSCOPY BASED ON STACKED AUTO-ENCODER

作者:李四海[1];余晓晖[2]

第一作者:李四海

机构:[1]甘肃中医药大学信息工程学院;[2]甘肃中医药大学药学院

第一机构:甘肃中医药大学信息工程学院(教育技术中心)

年份:2018

卷号:35

期号:6

起止页码:254

中文期刊名:计算机应用与软件

外文期刊名:Computer Applications and Software

收录:CSTPCD;;北大核心:【北大核心2017】;

基金:甘肃省自然科学基金项目(1506RJZA046);甘肃省中医药管理局项目(GZK-2013-44)

语种:中文

中文关键词:傅里叶变换红外光谱;深度学习;栈式自编码;定性分析

外文关键词:Fourier transform infrared spectroscopy (FTIR);Deep learning;Stacked auto-encode;Qualitative analysis

摘要:传统浅层模型不能有效提取FTIR光谱数据的潜在特征。提出一种基于栈式自编码器SAE(Stacked Auto Encoder)的光谱识别方法。通过堆叠稀疏自编码器构建深度网络,采用逐层贪婪训练学习光谱特征,根据学习到的特征有监督地训练softmax分类器,使用反向传播算法对网络进行微调。对麻花秦艽和大叶秦艽的FTIR光谱进行识别,基于SAE的分类准确率为96.67%,比偏最小二乘判别分析(PLSDA)和模型集群方法分别提高13.34%和10%。实验结果表明,该方法用于秦艽的快速、准确鉴别是可行和有效的。
The traditional shallow model failed to effectively extract the potential features of Fourier transform infrared(FTIR) spectroscopy. So a method is proposed for identification of FTIR spectroscopy based on stacked auto encoder(SAE). The deep network is constructed via stacking sparse auto encoders. Through the greedy layerwise approach,spectral characteristics are learned to train the softmax classifier with supervision,and the back-propagation algorithm is used to fine-tune the network parameter. The FTIR spectra of Gentiana macrophylla are identified and the accuracy reaches 96. 67%,which is 13. 34% and 10% higher than that of the partial least squares discriminant analysis(PLSDA) and model population method. The experimental results show that the proposed method is feasible and effective for rapid,accurate identification of Gentiana macrophylla.

参考文献:

正在载入数据...

版权所有©甘肃中医药大学 重庆维普资讯有限公司 渝B2-20050021-8 
渝公网安备 50019002500408号 违法和不良信息举报中心