一种用于强噪声环境下语音识别的含噪Lombard及Loud语音补偿方法

针对语音识别中由于强噪声的影响而引起的Lombard和Loud效应进行研究,提出了基于训练数据的加性噪声和Lombard及Loud效应的联合补偿法。对于加性噪声是从谱减法的逆向角度对训练数据在频谱域采用谱加法;对于Lombard和Loud语音,则采用基于隐马尔可夫模型(HMM)状态标注的训练数据补偿,该方法同时考虑Lombard和Loud语音不同声学单元的不同状态在倒谱域的多种变化和多种变异情况下不同声学单元的音长及相对音长的变化。这种基于数据的多模式补偿使模型自动适应多种噪声和语音变异情况,在强噪声环境下具有很强的鲁棒性,并且不影响识别系统在正常环境或正常发音时的识别性能.同时,由于补偿是在训练过程中得到,不增加识别时的计算复杂度。     

丁酮分子内价轨道1a″的电子动量谱学研究

用高分辨率电子动量谱仪进行丁酮分子的结合能谱和内价轨道1a″电子动量谱的实验工作，以及用Hartree-Fock和密度泛函理论方法对1a″轨道电子动量谱的理论研究. 得到了各价轨道的电离能值以及理论计算的总能、偶极矩和1a″轨道的二维密度图. 并比较了内价轨道1a″的电子动量谱的实验和理论计算结果，实验结果与理论计算符合较好.     

Monte Carlo simulations of down-scattered neutron and knock-on deuteron spectra in deuterium-tritium capsule implosions

A Monte Carlo particle tracking (MCPT) code has been developed and coupled to hydrodynamic simulations to generate and track primary and scattered neutrons in addition to scattered charged particles by post-processing. This code uses one dimensional (1-D) deuterium-tritium implosion profiles, and the computed down-scattered neutron and knock-on deuteron spectra are analyzed for different areal densities. The mixing effects on the spectra of down-scattered neutron and knock-on deuteron are also investigated. The implementation of the numerical scheme is analyzed, and the particle splitting technique is adopted, which is proven to efficiently reduce computational effort.     

Synthesis of C-Disaccharides An Unusual Ring Closure Reaction

Abstract

Reaction of the 4-C-lithiomethyl intermediates, obtained from 4-deoxy-4-C-iodomethylglucopyranosides 5a,b, with open-chain glucose derivative 8 afforded 4-deoxy-4-C-heptitylglucopyranosides 9a,b. Mesylation of the newly generated hydroxy group and then desilylation gave 11a,b which were subjected to ring closure under basic conditions. Surprisingly, no C-bridged cellobiosides 12a,b (or, alternatively the corresponding maltosides) were obtained as major products; with loss of one benzyl alcohol residue the furanosides 13a,b were preferentially formed. Their generation and structural assignment is discussed.     

Influence of structural defects on the optical properties of strongly coupled Au nanoshell arrays

The effects of different defects on optical properties and plasmon resonances properties of Au nanoshell arrays were investigated by using the finite-difference time-domain(FDTD) theory.It is found that the optical properties of the nanoshell arrays are strongly influenced by different defects.We show that when the hollow Au nanoshell arrays are placed in air,there is a wide photonic band gap(PBG) in the infrared region,but the band gap becomes narrower as we introduced different defects.Based on the distributions of electric field component E z and the total energy distribution of the electric and the magnetic field,we show that there exhibit dipoles field distributions for the plasmon mode at the long-wavelength edge of the band gap,but composite higher order modes are excited at the short-wavelength edge of the band gap.The plasmon resonant modes also can be controlled by introducing defects.     

Precise absolute positioning for a single-frequency user

A new strategy to realize precise absolute positioning for a single-frequency user is presented. In the presented strategy, the receiver clock and ambiguities are removed using the satelliteand epoch-differenced (SDED) algorithm. As a further development of the SDED algorithm, a regional augmentation network is used to generate the SDED atmospheric delays at the user. The weakened mathematic property due to the epoch-differenced operation is improved by adding the generated atmospheric delays and applying the robust estimation. To test the new approach, the 24-h data at 5 Continuous Operation Reference Station (CORS) stations in Shanghai is processed. The results show a more than 96% success rate, defined as the case where three directions achieve the desired positioning accuracy of 10 cm, when the observation is longer than 20 min. The 20-min static results show that the new method can reach an accuracy of 3.42, 4.76 and 9.26 cm in the North, East and Up directions, respectively. An experiment was carried out to assess the kinematic positioning. The results show that the kinematic positioning accuracy is 4.11, 5.31 and 4.05 cm in the north-south, east-west and height directions,respectively.     

晶体级联方式的宽带三倍频方案分析

 讨论了入射基频光频谱宽度、两块和频晶体失谐角对KDP晶体Ⅰ/Ⅱ/Ⅱ类角度失谐级联方式的三倍频光功率谱分布的影响,分析了影响三倍频效率的因素。研究结果表明,第二块和频晶体得到的三倍频光的功率谱分布与基频光功率谱分布及失谐角有关,三倍频光频谱宽度近似为基频光的3倍,采用该方案的三倍频可有效抑制由于基频光时间相位调制导致的三倍频光强随时间分布的不均匀,并能有效提高大带宽条件下的三倍频转换效率。同时,采用该方案的三倍频,对级联的两块和频晶体厚度及失谐角调整精度要求不高,在实验上有很大的可操作性。     

一种多源空间数据统一访问模型的实现

空间数据通常以栅格与矢量两种数据模型进行表达,但由于各自的不同特性,一般采用不同的访问方式.如何采用一致的方式进行多源空间数据访问是当前空间数据共享和互操作的难点之一.基于OGC SFS规范,从实现层次上考虑提出按几何与属性要素组织为几何-属性矢量数据模型（GAFeature model）,同时融入了扩展栅格数据模型,通过借鉴设计模式中的提供者模式和桥接模式,提供了一个较为完整的矢量栅格一体化访问管理模型,并进行了实现,使其能支持本地文件类型数据以及WCS、WFS空间信息服务.实验结果表明具备多源空间数据统一访问的能力.     

H2S在Fe(100)面吸附的第一性原理研究

基于密度泛函理论第一性原理, 在广义梯度近似下, 研究了表面覆盖度为0.25 ML (monolayer)时硫化氢分子在Fe(100)面吸附的结构和电子性质, 并与单个硫原子吸附结果进行了对比. 结果表明: 硫化氢分子吸附在B2位吸附能最小为-1.23 eV, 最稳定, B1位吸附能最大为-0.01 eV, 最不稳定; 并对硫化氢分子在B1位和B2位吸附后的电子态密度进行了分析, 也表明了吸附在B2位稳定, 且吸附在B2位后硫化氢分子几何结构变化不大; 将硫化氢中硫原子吸附与单个硫原子吸附的电子性质进行了比较, 发现前者吸附作用非常微弱; 同时对吸附后的Fe(100)面进行了对比, 单个硫原子吸附的Fe(100)面电子态密度出现了一系列峰值且离散分布, 生成了硫化亚铁, 表明在硫化氢环境下, 主要是硫化氢析出的硫原子发生了吸附. 关键词： 第一性原理 Fe(100)表面 吸附能 硫化氢     

pn结电容-电压法测量应变SiGe禁带宽度

利用应变SiGe/Si异质pn结电容-电压（C-V）特性确定SiGe禁带宽度的技术.该技术根据SiGe/Si异质pn结C-V实验曲线，计算出 pn结接触电势差，并得到SiGe/Si的价带偏移量和导带偏移量，进而求得SiGe禁带宽度.该技术测试方法简便，其过程物理意义清晰，既适用于分立的SiGe/Si异质pn结，也可直接分析SiGe/Si异质结器件中的SiGe 禁带宽度.实验结果与理论计算及其他相关文献报道的结果符合较好. 关键词： SiGe/Si 异质pn结 C-V 禁带宽度