基于短时傅立叶变换特征提取和卷积神经网络的LAMOST恒星光谱分类研究  

Stellar classification based on the convolutional neural network and short-time Fourier transform

作　　者：杜利婷 自彦丞 张静敏 艾霖嫔 周卫红[1,2] DU Li-ting;ZI Yan-cheng;ZHANG Jing-min;AI Lin-pin;ZHOU Wei-hong(School of Mathematics and Computer Science,Yunnan Minzu University,Kunming 650500,China;Key Laboratory of Celestial Structure and Evolution,Chinese Academy of Sciences,Kunming 650011,China)

机构地区：[1]云南民族大学数学与计算机科学学院,云南昆明650500 [2]中国科学院天体结构与演化重点实验室,云南昆明650011

出　　处：《云南民族大学学报（自然科学版）》2020年第5期480-485,共6页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：国家自然科学基金(61561053)。

摘　　要：光谱分类是研究恒星光谱的重要内容之一,对其进行准确分类识别在天文研究领域有着重要意义.提出一种新的光谱特征提取方法,利用短时傅里叶变换将一维光谱变换为二维傅里叶谱图像,然后利用卷积神经网络对得到的二维傅里叶谱图像进行分类,由于二维谱图像具有新的特征分布,提高了分类精度;在此基础上,为降低短时傅里叶变换中的采样过程造成的信息损失,在进行短时傅里叶变换前先利用一维卷积对一维恒星光谱数据进行处理,以提高分类准确率,实验结果显示证明了新的方法的有效性.Spectral classification is one of the important contents in the study of the stellar spectra,and it is of great significance to classify and recognize them accurately in the field of astronomy.A new spectral feature extraction method is proposed,in which a short-time Fourier transform is used to transform a one-dimensional spectral image into a two-dimensional Fourier spectral image,and then a convolutional neural network is used to classify the obtained two dimensional Fourier spectral images in order to reduce the information loss caused by the sampling process in the short-time Fourier transform.A new feature distribution of 2 D spectral images is proposed,which improves the classification accuracy.In order to improve the classification accuracy,one-dimensional Stellar Spectra data are processed by one-dimensional convolution before the short-time Fourier transform.The experimental results show that the new method is effective.

关 键 词：恒星光谱分类 卷积核 特征提取 短时傅里叶变换 卷积神经网络 

分 类 号：P144.1[天文地球—天体物理]