Tissue-specific δ13C in ancient and modern tropical seabirds and flying fish in the Xisha Islands,South China Sea

We analyzed ¹³C characteristics in samples of bird bones, feathers, eggshell carbonate and membrane from modern specimens of red-footed booby (Sula sula) as well as fish muscle, scales, and bones from its predominant food source, flying fish (Exocoetus volitans), and muscle from its secondary food source squid (Loligo chinensis), as well as in ancient sub-fossil samples of seabird and flying fish at the Xisha Islands, South China Sea. δ¹³C is tissue-specific in both seabirds and flying fish due to the variance in turn-over among the tissues and differences in the type and content of amino acids across a diverse range of tissues. The δ¹³C discrimination factors also differed significantly among the various tissues between tropical seabirds and their prey. A Suess effect, caused by fossil fuel combustion and the emission of carbon with fewer ¹³C isotopes, was observed in the bird and fish tissue from ancient to modern time. Our study provides a multiple variability index for δ¹³C in organisms along a food chain, and verifies that tissue-specific ¹³C analysis is essential to identify diet and species and thus is a valuable tool for research on tropical seabird ecology.     

基于卷积神经网络的烟叶近红外光谱分类建模方法研究

卷积神经网络(CNN)在图像分类识别领域应用广泛,但其在近红外光谱分类中的研究还未见报道,对基于CNN的近红外光谱分类建模方法进行了研究。针对近红外光谱数据的特点,提出了一种改进的卷积神经网络建模方法,对CNN经典模型LeNet-5所做改进：①将方形矩阵卷积核改为适用于一维近红外光谱的向量卷积核;②简化网络结构,将LeNet-5结构中C5,F6及输出层改为单层感知机。同时,采用隔点采样的方法对近红外光谱降维,加快收敛速度;并对卷积核尺寸对建模结果的影响进行了研究。以我国东北、黄淮、西南三大烤烟产区的600个中部烟叶样本的近红外光谱为实验对象,建立烟叶产区分类NIR-CNN模型。该模型对训练集和测试集的判别准确率为98.2%和95%。实验结果表明,应用卷积神经网络可对近红外光谱数据准确、可靠地判别分类;烟叶产区NIR-CNN建模方法可为卷烟企业烟叶原料科学合理利用提供指导,为维护卷烟产品的质量稳定有重要意义;基于卷积神经网络的近红外光谱判别方法也可推广到其他农产品的分类应用中。     

稀疏降噪自编码结合高斯过程的近红外光谱药品鉴别方法

提出一种稀疏降噪自编码结合高斯过程的近红外光谱药品鉴别方法。首先对近红外光谱数据进行小波变换以消除基线漂移,然后用稀疏降噪自编码(SDAE)网络提取光谱特征并降维表示,最后采用高斯过程(GP)进行二分类,其中GP选用光谱混合(SM)核函数作为协方差函数,记此分类网络为wSDAG_SM。自编码网络具有很强的模型表示能力,高斯过程分类器在处理小样本数据时具有优势。wSDAG_SM网络通过稀疏降噪自编码学习得到维数更低但更有价值的特征来表示输入数据,同时将具有很好表达力的光谱混合核作为高斯过程的协方差函数,有利于更准确的光谱数据分类。以琥乙红霉素及其他药品的近红外光谱为实验数据,将该方法与经过墨西哥帽小波变换的BP神经网络(wBP)、支持向量机(wSVM), SDAE结合Logistic二分类(wSDAL)、SDAE结合采用平方指数(SE)协方差核的GP二分类(wSDAG_SE),以及未采用小波变换的SDAG_SM网络等方法进行对比。实验结果表明,对光谱数据进行墨西哥帽小波变换预处理能有效提升SDAG_SM网络的分类准确率和稳定性。wSDAG_SM方法无论从分类准确率还是分类结果稳定性方面,都优于其他分类器。     

Fisher判别在气相色谱-质谱分析助燃剂燃烧残留物中的应用

为寻找一种用于火场助燃剂燃烧残留物鉴定的更为准确、有效的模式识别方法,对7种常见助燃剂在不同载体上的燃烧残留物样品及未知送检样品进行气相色谱-质谱（GC-MS）分析测试,通过特征组分分析鉴定出未知样品中含有汽油成分。同时运用Fisher判别及PCA（主成分分析）/Fisher判别联用两种判别方法对样本数据进行了分析处理,PCA/Fisher判别联用的结果表明送检样本中含有硝基油漆稀料成分,而仅使用Fisher判别的结果表明送检样本中含有93^#汽油。通过将两种分析方法所得结果与GC-MS特征组分分析的结果进行比对发现,Fisher判别能够对7种助燃剂燃烧残留物的样本实现更有效的分类,对未知样本的判别更为有效。该研究结果为火场助燃剂鉴定提供了新的数据分析手段。     

基于无机元素分析对地理标志五常大米鉴别技术的研究

五常大米是我国著名地理标志产品之一,品质高、产量低,导致制假售假的现象严重。为维护五常大米的品牌形象和消费者利益,急需一种有效鉴别五常大米的方法。应用电感耦合等离子体光谱及电感耦合等离子体质谱测定大米中无机元素含量,结合主成分分析(PCA)、Fisher判别、人工神经网络(ANN)对五常大米鉴别模型进行研究。结果表明：PCA法对样品的分类效果较差,采用Fisher判别和ANN则可准确识别五常地区的大米样品和其他地区的大米样品。Fisher判别法对校正集和验证集样品平均准确识别率分别为93.5%,而ANN法对同样的校正集和验证集样品的平均准确识别率为96.4%,优于Fisher判别法。可准确对五常大米进行鉴别,为该产品的地理标志保护提供了一种技术手段。     

Plastic scintillator for pulse shape neutrons and gamma quanta discrimination

This study describes a new plastic scintillator for pulse shape n-γ discrimination. The scintillator contains two activation centers with different life spans. The first activator collects the singlet excitation energy of a polymer base, and the second activator utilizes triplet excitation states. We utilized 1,4-dimethyl-9,10-diphenylanthracene (DMDPA) and tris(dibenzoylmethide) (1,10-phenanthroline)Europium(III) (Eu[DBM]₃Phen) as activators. The figure of merit for this scintillator is 1.37, which is sufficient for reliable n-γ discrimination.     

近红外光谱技术对闽南乌龙茶品种的识别研究

采用近红外光谱技术建立了一种快速无损的乌龙茶品种识别方法。收集闽南地区不同茶场中铁观音、黄金桂、本山、毛蟹与梅占等5个品种共210份具有代表性的乌龙茶样品,采集近红外光谱数据,选用1 100~1 300nm,1 640~2 498nm作为检测波长范围,利用主成分分析法(principal component analysis,PCA)建立模型,并在实验过程中比较多元散射校正(multiplicative scatter correction,MSC)与标准正态变量校正(standard normal variate,SNV)两种数据预处理方法对模型的影响。实验结果表明,多元散射校正对模型的影响优于标准正态变量校正,对校正集的识别准确率达到了96%,对预测集中样品的识别准确率达到了90%。实验结果证明了采用近红外光谱技术可以快速无损识别闽南地区乌龙茶,具有较强的实用价值和推广价值。     

有距离辨识度的传声器阵列波束优化算法研究*

小尺度线性传声器阵列在便携式通信设备中具有潜在应用前景,其波束优化算法大多关注于抑制来自与声源不同方向的噪声,而不能有效地抑制与声源同方向的噪声。本文通过对几种典型波束优化策略进行改进,增加线性约束条件,使得阵列兼具方向和距离辨识度。论文对比了几种优化方法的波束性能,同时讨论了约束条件对波束旁瓣的影响,并提出了鲁棒性的优化方案。仿真结果显示,合理的优化策略和线性约束条件可以使阵列波束获得很好的聚焦性能。     

Investigation of brush-type chiral stationary phases based on O,O'-diaroyl tartardiamide and O,O'-bis-(arylcarbamoyl) tartardiamide

Four chiral organosilanes based on O,O'-dibenzoyl tartardiamide, O,O'-bis-(3,5-dimethylbenzoyl) tartardiamide, O,O'-bis-(phenylcarbamoyl) tartardiamide and O,O'-bis-[(3,5-dimethylphenyl)carbamoyl] tartardiamide were synthesized and immobilized on silica to afford corresponding brush-type chiral stationary phases (CSPs) with well-defined structures. Using 54 compounds containing a wide variety of structures as analytes, the enantioselectivities of the four CSPs were evaluated under normal-phase modes. 3,5-Dimethyl substituent in the aryl group was found to significantly affect the enantioselectivity of CSPs containing aryl ester moieties. Aryl carbamate moieties in CSPs were observed more beneficial for enantioseparation than aryl ester moieties. The additional hydrogen-bond donors (NH) present in the carbamate groups contributed greatly to the enantioselectivity of CSPs, which is contrary to the results that have been found in network-polymeric CSPs.