偏度最大化多通道逆滤波语声去混响研究

房间混响会降低语声质量和语声可懂度。高阶统计量是衡量非高斯性的重要参量,基于语声非高斯特性可实现语声去混响。该文提出一种基于高阶统计量的多通道语声去混响方法,该方法首次用多通道语声信号线性预测残差的三阶统计量偏度构造代价函数,以去混响重建信号线性预测残差的偏度最大化为目标自适应地更新逆滤波器,同时引入通道逆滤波和语声产生系统的联合估计。实验结果表明,该方法相较于已有的基于线性预测残差四阶统计量峰度的方法具有更好的去混响效果,且对噪声具有更强的鲁棒性。     

应对说话人位置突变的鲁棒语声去混响*

多通道线性预测算法是一种有效的语声去混响方法,但若房间内说话人位置发生变化,房间冲激响应也会随之改变,从而导致去混响效果变差。本文研究了基于递归最小二乘算法的自适应多通道线性预测语声去混响方法,提出了一种去混响过程中检测说话人位置突变的方法,它借助已有的多通道传声器信号,并未明显增加系统开销。实验表明该方法可以有效检测说话人位置的突变,并通过重置滤波器改善算法的收敛速度和稳定性。     

球谐域自适应混响抵消与声源定位算法

提出了一种基于球谐域的自适应混响抵消与声源定位算法,该方法通过去混响处理改善语音质量,并提高球谐域定位算法在混响环境下的定位性能。推导了基于多通道线性预测的自适应混响抵消算法在球谐域的表达式,针对刚球模型提出分阶处理的去混响方法,并对去混响后的信号进行波达方向估计。采用32元球阵的仿真结果表明,相比于球谐域不分阶去混响方法,该方法最大可减少约2/3的运算量,同时语音PESQ得分及SRMR均显著提高。利用实验数据对算法性能进行测试,实验结果验证了该方法在实际声学环境中去混响和声源定位的有效性。      

基于卡尔曼滤波的低复杂度去混响算法*

在电话会议、智能音箱等应用场景下,传声器往往处在声源的远场。混响信号的存在会掩蔽后续到达的直达声信号,降低传声器接收信号的语音质量,以及语音识别系统的准确识别率。多通道线性预测算法是一种经典的盲去混响算法,但该算法往往具有较高的计算复杂度。本文提出了一种简化的卡尔曼滤波更新算法,通过对角化卡尔曼滤波器状态向量误差协方差矩阵,降低了自适应多通道线性预测去混响算法的复杂度。通过与现有分块对角简化算法对比发现,本文提出的简化算法在保证语音质量的同时,进一步降低了原卡尔曼滤波算法的复杂度。     

n型金属氧化物半导体场效应晶体管噪声非高斯性研究

基于n型金属氧化物半导体场效应晶体管（nMOSFET）噪声的数涨落模型,采用高阶统计量双相干系数平方和研究了nMOSFET噪声的非高斯性.通过对nMOSFET实际测试噪声的分析,发现nMOSFET器件噪声存在非高斯性;小尺寸器件噪声的非高斯性强于大尺寸器件;在器件的强反型线性区,其非高斯性随着漏压的增加而增加.文中还通过蒙特卡罗模拟和中心极限定理理论对nMOSFET噪声的非高斯性作了深入的探讨. 关键词： 噪声 非高斯性 n型金属氧化物半导体场效应晶体管 氧化层陷阱     

线性非高斯序列的替代数据检验新方法

替代数据法作为检验时间序列非线性和混沌的统计方法获得了广泛的应用.由于原替代数据法的零假设为线性高斯过程,可能把线性非高斯过程,特别是非最小相位过程误判为非线性.为了解决这一问题,提出并详细推导了基于功率谱等价的非最小相位序列求逆方法;结合基于高阶累积量的非最小相位自回归滑动平均模型辨识方法,提出了检验序列是否为线性非高斯过程的替代数据生成新算法.仿真算例表明,上述方法成功地克服了原替代数据法的不足. 关键词： 替代数据 非线性检验 非最小相位 功率谱等价     

考虑帧间信息的语音带宽扩展

语音带宽扩展是为了提高语音质量,利用语音低频和高频之间的相关性重构语音高频的一种技术。高斯混合模型法是语音带宽技术中被广泛应用的一种方法,但是,该方法的映射函数是分段线性函数,且没有考虑语音前后帧的相关信息。因此,提出了一种基于条件受限玻尔兹曼机的方法。该方法利用条件受限玻尔兹曼机提取了语音信号的帧间信息,同时将语音低频、高频特征参数映射为高阶统计特性,深层发掘和模拟了语音低频和高频之间的非线性关系。客观和主观对比测试结果都表明,该方法性能优于传统的高斯混合模型方法。      

一种频域合成房间频率响应的人工混响方法

给出了一种频域合成房间频率响应的方法用于卷积法人工混响,基于频域内房间频率响应的后期部分为高斯随机过程的假设,用自回归滑动平均模型为其自协方差函数和功率谱密度进行参数化描述,在对自回归滑动平均模型中的参数求解后,通过逆滤波得到了房间频率响应后期部分,与房间频率响应前期部分组合后经过傅里叶反变换得到完整的房间脉冲响应。仿真结果表明该方法的混响效果与镜像源法接近,明显优于反馈延迟网络法,但其计算复杂度比镜像源法低,便于实时应用。     

频域合成房间频率响应的人工混响方法

给出了一种频域合成房间频率响应的方法用于卷积法人工混响,基于频域内房间频率响应的后期部分为高斯随机过程的假设,用自回归滑动平均模型为其自协方差函数和功率谱密度进行参数化描述,在对自回归滑动平均模型中的参数求解后,通过逆滤波得到了房间频率响应后期部分,与房间频率响应前期部分组合后经过傅里叶反变换得到完整的房间脉冲响应。仿真结果表明该方法的混响效果与镜像源法接近,明显优于反馈延迟网络法,但其计算复杂度比镜像源法低,便于实时应用。     

改进的噪声鲁棒语音稀疏线性预测算法

语音线性预测分析算法在噪声环境下性能会急剧恶化,针对这一问题,提出一种改进的噪声鲁棒稀疏线性预测算法。首先采用学生t分布对具有稀疏性的语音线性预测残差建模,并显式考虑加性噪声的影响以提高模型鲁棒性,从而构建完整的概率模型。然后采用变分贝叶斯方法推导模型参数的近似后验分布,最终实现噪声鲁棒的稀疏线性预测参数估计。实验结果表明,与传统算法以及近几年提出的基于l_1范数优化的稀疏线性预测算法相比,该算法在多项指标上具有优势,对环境噪声具有更好的鲁棒性,并且谱失真度更小,因而能够有效提高噪声环境下的语音质量。     

Speckle characteristics of a broad-area VCSEL in the incoherent emission regime

We present a technique to reduce the speckle contrast of a NIR broad-area VCSEL based on the spatially incoherent emission regime that can be obtained when using the proper driving conditions. We evaluate the efficiency of this technique to reduce the speckle contrast by comparing it with the speckle characteristics in multimode emission under cw operation. Depending on the illumination setup, the incoherent emission regime can lead to a strongly reduced speckle contrast down to 1.3%. This is in agreement with estimates of the expected speckle contrast reduction when three contrast reducing effects are taken into account. These low contrast values make the investigated sources attractive for several applications that suffer from speckle noise.     

XYZ particles at Belle

In this paper, I review recent progress in the study of the XYZ particles at Belle. I only focus on studies with charmonium and one or more light mesons in the final states. This covers the X(3872), X(3915),Y(4140), X(4350), and the charged Z states.     

Beyond the LAN: Ethernet’s Evolution into the Public Network

This article details the evolution of Ethernet into Gigabit Ethernet and how this LAN-based technology has undergone major transformations over time. From its data rates and distances to supported media and functionality, Ethernet has greatly improved, enabling it to surmount many of its former limitations and in so doing to expand beyond the LAN into the MAN and now even the WAN. In this article, Pioneer Consulting explores the evolution further by focusing on some of the major technological directions in the Ethernet equipment industry.     

Optical transmission through an interface between subwavelength slits

This paper studies optical transmission through an interface between two slits with different widths in a sheet composed of an ideal conductor. Such a structure is of potential use in fabricating optical diode and may be the simplest one compared to other designs. Our calculations show that there is a critical wavelength. When the light wavelength is below the critical wavelength, the transmissivity is unidirectional. The expression of the stable transmissivity as a function of the ratio of the widths of the two slits was obtained analytically. Particularly, at the critical wavelength, the transmissivities are zero. This phenomenon has great potential for application in the manufacture of wavelength blockers.     

Effect of swift heavy ion irradiation on photoluminescence properties of ZnO/PMMA nanocomposite films

We report a study on the SHI induced modifications on structural and optical properties of ZnO/PMMA nanocomposite films. The ZnO nanoparticles were synthesized by the chemical route using 2-mercaptoethanol as a capping agent. The structure of ZnO nanoparticles was confirmed by XRD, SEM and TEM. These ZnO nanoparticles were dispersed in the PMMA matrix to form ZnO/PMMA nanocomposite films by the solution cast method. These ZnO/PMMA nanocomposite films were then irradiated by swift heavy ion irradiation (Ni⁸⁺ ion beam, 100 MeV) at a fluence of 1×10¹¹ ions/cm². The nanocomposite films were then characterized by XRD, UV-vis absorption spectroscopy and photoluminescence spectroscopy. As revealed from the absorption spectra, absorption edge is not changed by the irradiation but the optical absorption is increased. Enhanced green luminescence at about 527 nm and a less intense blue emission peak around 460 nm were observed after irradiation with respect to the pristine ZnO/PMMA nanocomposite film.     

带有碳杂质的钨中氢稳定性的第一性原理研究

应用第一性原理计算方法研究了碳(C)原子对钨(W)中氢(H)原子稳定性的影响. 本征W中, 当C-H间距离为～2.5 Å时, H的溶解能出现最低值, 此时为H最稳定的位置. W中存在空位时, 由于C的影响, H占据的最佳电子密度面值为0.10 Å^-3. 研究发现, W中单空位最多能容纳10个H原子, 且不能形成H分子, 不同于没有C存在的情况, 表明C对W中H稳定性存在很大影响. 此外, 当两个C原子存在于空位中时, H占据的最佳电子密度面值变为0.13 Å^-3.     

Grain boundary potential and charge density at grain boundaries of chemically prepared thin silver sulfide films

The grain boundary potential and interface state charge density at the grain boundaries of silver sulfide (Ag₂S) thin films prepared by chemical conversion of cadmium sulfide (CdS) films have been determined from the dc resistance of the material and are found to be sensitive to annealing. A reduction in the grain boundary potential and the grain boundary charge density of the film has been noticed when the source CdS film is annealed at different temperatures prior to chemical conversion. The variation in the grain boundary charge density of the grown Ag₂S film with source annealing temperature has been found to be similar to that of thin cadmium sulfide film, reported earlier. An additional low temperature heat treatment of the sample results in an enhancement in the charge density at the grain boundaries. The change in the silver vacancy and/or oxygen and sulfur content of the films as revealed from the energy dispersive spectra of the films suggests possible role of film composition on the grain boundary charge density.     

Reflections from polished single mode fiber ends

Abstract

The performance of a wavelength division multiplexing (WDM) system with optical cross-connects (OXCs) and in the presence of polarization and nonlinear effects was analyzed. We determined the effect of PMD and PDL; nonlinearity such as XPM, SPM, and FWM; and CD on the system performance. We found that the system outage probability is highest at DGD threshold for the respective maximum eye closure (EC). We showed that the allowable DGD depends on the time slot for one bit T, which is the inverse of the network bit rate. Wider T is more immune to large PMD with the EC still maintained at the same level since DGD/T is unchanged. Our analysis also indicated that as DGD increases, BER increases. The BER also increases with increase in PDL and further increases with the presence of XPM, SPM, FWM, and CD.     

The Effects of Polarization and Non-Linearity on the Performance of WDM System with Optical Cross-Connects

The performance of a wavelength division multiplexing (WDM) system with optical cross-connects (OXCs) and in the presence of polarization and nonlinear effects was analyzed. We determined the effect of PMD and PDL; nonlinearity such as XPM, SPM, and FWM; and CD on the system performance. We found that the system outage probability is highest at DGD threshold for the respective maximum eye closure (EC). We showed that the allowable DGD depends on the time slot for one bit T, which is the inverse of the network bit rate. Wider T is more immune to large PMD with the EC still maintained at the same level since DGD/T is unchanged. Our analysis also indicated that as DGD increases, BER increases. The BER also increases with increase in PDL and further increases with the presence of XPM, SPM, FWM, and CD.     

化学发光法测定巯嘌呤的含量

在流动注射系统中,利用巯嘌呤在碱性Luminol-K₃Fe(CN)₆化学发光体系中发光信号强的特点,建立了一种测定巯嘌呤的新方法。在优化条件下,方法的线性范围是1.02×10-7～1.02×10-9mol·L-1, 回归方程是Y=36.315c+140.72(Y是相对发光强度,c是巯嘌呤浓度与10-8 mol·L-1的乘积), 相关系数是0.998 2,检出限(S/N=3)是6.33×10-10mol·L-1,RSD为3.54%（c_i=2.0×10-9mol·L-1, n=12）。该法用于合成样品的测定,简便、灵敏、快速,结果令人满意。文章简要地探讨了巯嘌呤在碱性Luminol-K₃Fe(CN)₆化学发光体系中的化学发光机理：在反应过程中,产生大量的过氧自由基和羟基自由基,自由基与鲁米诺反应,形成激发态的鲁米诺分子,鲁米诺分子由激发态回到基态产生化学发光现象。