空间波函数的对称性对氦原子双光子双电离过程中电子关联的影响

通过数值求解含时薛定谔方程，研究了氦原子具有对称空间波函数的1s2s 1S态和具有反对称空间波函数的1s2s 3S态分别作为初态的双光子双电离过程. 结果表明，对于初态为单重态1s2s 1S的双光子双电离过程，两个电离电子的能量分布随激光脉冲持续时间的增加呈现由单峰到双峰的变化，这里的单峰和双峰分别意味着两个电离电子主要携带相等和不等的能量；然而对于初态为三重态1s2s 3S的双光子双电离过程，两个电离电子的能量分布随激光脉冲持续时间的增加总是保持双峰结构. 这些结果表明当原子的初态处于反对称空间波函数时，两电子的空间密度分布具有较少的重叠，从而导致电子在超短激光脉冲中电离时电子关联能无法平均分配.     

Seventeen‐Coordinate Actinide Helium Complexes

The geometries and electronic structures of molecular ions featuring He atoms complexed to actinide cations are explored computationally using density functional and coupled cluster theories. A new record coordination number is established, as AcHe₁₇³⁺, ThHe₁₇⁴⁺, and PaHe₁₇⁴⁺ are all found to be true geometric minima, with the He atoms clearly located in the first shell around the actinide. Analysis of AcHe_n ³⁺ (n =1–17) using the quantum theory of atoms in molecules (QTAIM) confirms these systems as having closed shell, charge‐induced dipole bonding. Excellent correlations (R ²>0.95) are found between QTAIM metrics (bond critical point electron densities and delocalization indices) and the average Ac−He distances, and also with the incremental He binding energies.     

汉语儿童情感语声合成

情感语声合成技术对于人机交互具有重要的意义。面对儿童情感语声合成所需汉语语声数据资源缺乏以及模型训练时长较长等问题,该文提出利用迁移学习实现汉语儿童情感语声合成的方法。首先基于汉语语声数据库训练深度学习模型实现中文语声端到端合成模型,再使用高质量大样本的中文情感语料库完成情感语声合成模型,最后利用自行采样的小样本汉语儿童情感语料对模型进行迁移学习实现低资源的语声合成。客观实验结果中梅尔倒谱失真指标为4.91,主观听辨实验指标分别为3.61和4.17。通过实验对比表明,该文的方法在情感语声合成技术的应用上具有良好的性能表现,并且优于现有先进的低资源情感语声合成方法。     

Att-U-Net：融合注意力机制的U-Net骨导语声增强*

近年来大量全卷积网络、U-Net等编解码网络结构应用于语音增强,它们具有计算复杂度低、模型参数少等优势。然而,与长短时记忆模型等方法相比,这些编解码结构仍存在不能充分利用先后时间之间和高低频率之间的关联信息等缺点,尤其对于长序列数据的输入,编解码结构存在信息丢失的问题。为保持计算效率的同时考虑更充分的时频关联信息建模,本文提出一种融合注意力机制的U-Net网络的骨导语音增强方法（Att-U-Net）,通过在跳跃连接中引入注意力机制,生成一个权重矩阵,将编码层中的全局信息根据权重融入对应的解码层中,使网络在编解码过程中能够关注输入数据中与增强目标相关程度高的重要信息,同时抑制不相关的信息。在骨导语音数据集上的实验表明,融合注意力机制的U-Net网络能在保持模型轻量化的同时有效提升骨导语音的增强效果,增强后的语音在各项客观评价指标上均优于基线模型。通过对编解码网络中间层的可视化分析发现,在解码过程中注意力机制有效地保留了有声段的信息,滤除了骨导语音由于骨导传声特性带来的中频共振,从而使得增强后的骨导语音具有较好的听觉效果。     

Visual Speech Recognition with Lightweight Psychologically Motivated Gabor Features

Extraction of relevant lip features is of continuing interest in the visual speech domain. Using end-to-end feature extraction can produce good results, but at the cost of the results being difficult for humans to comprehend and relate to. We present a new, lightweight feature extraction approach, motivated by human-centric glimpse-based psychological research into facial barcodes, and demonstrate that these simple, easy to extract 3D geometric features (produced using Gabor-based image patches), can successfully be used for speech recognition with LSTM-based machine learning. This approach can successfully extract low dimensionality lip parameters with a minimum of processing. One key difference between using these Gabor-based features and using other features such as traditional DCT, or the current fashion for CNN features is that these are human-centric features that can be visualised and analysed by humans. This means that it is easier to explain and visualise the results. They can also be used for reliable speech recognition, as demonstrated using the Grid corpus. Results for overlapping speakers using our lightweight system gave a recognition rate of over 82%, which compares well to less explainable features in the literature.     

语音信号序列的Volterra预测模型

对给定的英语音素、单词和语句进行了采集并完成预处理. 分别应用互信息法和Cao 氏法确定了实际采集的语音信号序列的延迟时间和嵌入维数, 以完成语音序列的相空间重构. 通过计算实际采集的语音信号序列的最大Lyapunov指数, 完成了语音信号的混沌特性识别, 判定其具有混沌特性. 引入Volterra级数, 提出了一种具有显式结构的语音信号非线性预测模型. 为克服最小均方误差算法在Volterra模型系数更新时固有的缺点, 在最小二乘法基础上, 应用基于后验误差假设的可变收敛因子技术, 构建了一种基于Davidon-Fletcher-Powell算法的二阶Volterra 模型(DFPSOVF), 并将其应用于具有混沌特性的语音信号序列预测. 仿真结果表明: DFPSOVF非线性预测模型对于单帧和多帧语音信号均具有更好的预测精度, 优于线性预测模型, 并且能够很好地反映语音序列变化的趋势和规律, 完全可以满足语音预测的要求; 可以根据语音信号序列的嵌入维数选取预测模型的记忆长度. 所提出模型可以为语音信号重构和压缩编码开辟一条新途径, 以改善语音信号处理方法的复杂度和处理效果.     

Chemical characterization and ultrastructure study of pulp fibers

Understanding the ultrastructure and chemical characterization of pulp fibers is highly important in utilizing wood as a raw material in a wide scope of applications, such as forest biomass-based biorefineries and low-cost renewable materials. The observation of the ultrastructure is not possible without advanced microscopy and spectroscopy techniques. Therefore, this study focuses on exploring the ultrastructure of pulp fibers with helium ion microscopy (HIM) and scanning electron microscopy (SEM). For the analysis of chemical characterization in the pulp fibers, Raman spectroscopy, Fourier transform infrared spectroscopy (FT-IR), and X-ray photoelectron spectroscopy (XPS) were performed. For these studies, the pulp fiber samples were obtained mainly from three different wood species, i.e. spruce, birch and eucalyptus. They were received in the never dried state and dried with a critical point drier (CPD) to minimize pore collapse. The spectroscopy results showed a strong signal from crystalline cellulose and confirmed the absence of lignin after Kraft pulping and bleaching. However, with XPS about 2% of lignin was detected in eucalyptus pulp. The results obtained with the microscopy techniques are compared and indicating the nanofibril size, shape, surface roughness as well as their orientation in pulp fibers. To our knowledge, this is the first time that HIM is applied to study the ultrastructure of pulp fibers and compared against more conventional microscopy and spectroscopy techniques. The main differences between HIM and SEM were found to be related to the focusing and magnification. The individual nano- and microfibrils as well as their bundles were more easily visible with HIM than with SEM. Also, with HIM it was possible to get the total area in focus at once which was not the case with SEM. The increased understanding of the ultrastructure and chemical composition of wood pulp enhance the development of novel wood-based products and processes for their manufacture.     

氦原子基态能量的组态相互作用计算

将基态氦原子的波函数取作1s2和1s2s两个组态函数的叠加,利用组态相互作用方法解析计算了氦原子基态的非相对论能量.计算结果表明,考虑激发态1s2s与基态的相互作用,可以获得0.029 38Hartree的基态能量修正.本文的解析组态相互作用方法可作为量子力学教学的有益补充.