各类大型巡天项目产生了海量的天文数据,因此,需要研究适用于大规模数据的光谱自动处理方法.传统的基于谱线检测或BPT(Baldwin,Phillips,Terjevich)诊断图的星系光谱分类方法难以直接应用于星系光谱自动分类,相比之下,基于机器学习的光...各类大型巡天项目产生了海量的天文数据,因此,需要研究适用于大规模数据的光谱自动处理方法.传统的基于谱线检测或BPT(Baldwin,Phillips,Terjevich)诊断图的星系光谱分类方法难以直接应用于星系光谱自动分类,相比之下,基于机器学习的光谱自动分析更适用于海量天文数据的分类研究.提出一种基于双层聚类的星系光谱分析方法.第1层采用k均值聚类算法将星系光谱分为吸收线星系和发射线星系,第2层使用CLARA(Clustering LARge Applications)聚类算法将发射线星系聚为5簇.对LAMOST DR5的星系数据进行实验,结果表明:(1)第1层k均值聚类能够成功将星系光谱分为吸收线星系和发射线星系,聚类簇与基于谱线检测的分类结果基本一致.(2)第2层CLARA聚类结果能够在BPT图中反映出不同的星系类型.(3)光谱聚类结果与颜色星等图分类存在预期的相关性.(4)k均值聚类和CLARA聚类能够适用于大规模数据自动分析处理,聚类结果能够很好地反映星系的物理性质和演化过程,簇心数据可以为光谱自动分类系统提供模板.展开更多
To make the modulation classification system more suitable for signals in a wide range of signal to noise ratios (SNRs), a novel adaptive modulation classification scheme is presented in this paper. Differ-ent from ...To make the modulation classification system more suitable for signals in a wide range of signal to noise ratios (SNRs), a novel adaptive modulation classification scheme is presented in this paper. Differ-ent from traditional schemes, the proposed scheme employs a new SNR estimation algorithm for small samples before modulation classification, which makes the modulation classifier work adaptively according to estimated SNRs. Furthermore, it uses three efficient features and support vector machines (SVM) in modulation classification. Computer simulation shows that the scheme can adaptively classify ten digital modulation types (i.e. 2ASK, 4ASK, 2FSK, 4FSK, 2PSK, 4PSK, 16QAM, TFM, π/4QPSK and OQPSK) at SNRS ranging from 0dB to 25dB and success rates are over 95% when SNR is not lower than 3dB. Accuracy, efficiency and simplicity of the proposed scheme are obviously improved, which make it more adaptive to engineering applications.展开更多
文摘各类大型巡天项目产生了海量的天文数据,因此,需要研究适用于大规模数据的光谱自动处理方法.传统的基于谱线检测或BPT(Baldwin,Phillips,Terjevich)诊断图的星系光谱分类方法难以直接应用于星系光谱自动分类,相比之下,基于机器学习的光谱自动分析更适用于海量天文数据的分类研究.提出一种基于双层聚类的星系光谱分析方法.第1层采用k均值聚类算法将星系光谱分为吸收线星系和发射线星系,第2层使用CLARA(Clustering LARge Applications)聚类算法将发射线星系聚为5簇.对LAMOST DR5的星系数据进行实验,结果表明:(1)第1层k均值聚类能够成功将星系光谱分为吸收线星系和发射线星系,聚类簇与基于谱线检测的分类结果基本一致.(2)第2层CLARA聚类结果能够在BPT图中反映出不同的星系类型.(3)光谱聚类结果与颜色星等图分类存在预期的相关性.(4)k均值聚类和CLARA聚类能够适用于大规模数据自动分析处理,聚类结果能够很好地反映星系的物理性质和演化过程,簇心数据可以为光谱自动分类系统提供模板.
文摘To make the modulation classification system more suitable for signals in a wide range of signal to noise ratios (SNRs), a novel adaptive modulation classification scheme is presented in this paper. Differ-ent from traditional schemes, the proposed scheme employs a new SNR estimation algorithm for small samples before modulation classification, which makes the modulation classifier work adaptively according to estimated SNRs. Furthermore, it uses three efficient features and support vector machines (SVM) in modulation classification. Computer simulation shows that the scheme can adaptively classify ten digital modulation types (i.e. 2ASK, 4ASK, 2FSK, 4FSK, 2PSK, 4PSK, 16QAM, TFM, π/4QPSK and OQPSK) at SNRS ranging from 0dB to 25dB and success rates are over 95% when SNR is not lower than 3dB. Accuracy, efficiency and simplicity of the proposed scheme are obviously improved, which make it more adaptive to engineering applications.