稀疏低秩噪声模型下无监督实时单通道语音增强算法

针对现有基于字典学习的增强算法需要先验信息、不易实时处理的问题,提出一种便于实时处理的无监督的单通道语音增强算法。首先,该算法将无监督条件下背景噪声的建模问题转化为带噪语音幅度谱的稀疏低秩噪声分解;然后,采用增量非负子空间方法对背景噪声进行在线字典学习,获得能够体现背景噪声时变特性的自适应噪声字典;最后,利用所得的噪声字典,采用易于实时处理的逐帧迭代方式,对带噪语音进行处理。实验结果表明:相较于多带谱减法和基于低秩稀疏矩阵分解的增强算法,所提算法在噪声抑制方面的性能尤为显著,在多项性能评价指标上,均表现出更好的结果。     

Mo低掺杂La0.67Sr0.33MnO3的磁电特性及室温磁电阻增强

采用固相反应法成功制备了Mo低掺杂La0.67Sr0.33Mn1-xMoxOc(x=0,0.02)单相多晶样品,系统研究了样品的磁性、电输运特性和磁电阻效应.发现Mo低掺杂对La0.67Sr0.33MnO3 的居里温度影响很小,但在La0.67Sr0.33MnO3铁磁基态背景下诱导出团簇自旋玻璃态行为.在低温区(0.1相似文献   

Reinvestigation of the high spin states in ^（161）Er and enhanced E1 transitions in the N=93 isotones

High-spin states in ¹⁶¹Er have been studied experimentally using the ¹⁵⁰Nd(¹⁶O, 5n) reaction at a beam energy of 86 MeV. The relatively enhanced E1 transitions between the 5/2⁺[642] and 3/2^-[521] bands are observed in ¹⁶¹Er, and the B(E1) values are extracted experimentally. The systematics of the R(E1) values in the N=93 isotones are presented. It is found that the strength of the E1 transitions obviously exhibits angular momentum dependence, and the occurrence of the relatively enhanced E1 transitions could be attributed to octupole softness.     

二氧化硅纳米结构径向呼吸振动的理论研究

运用密度泛函理论的B3LYP方法,在6-31G(d)基组水平上,计算了由Si₂O₂二元环组成的准一维SiO₂纳米链及三元环(3MR)、四元环(4MR)、六元环(6MR)和八元环(8MR)组成的SiO₂纳米管的Raman振动光谱。从Raman光谱发现纳米链和纳米管的径向呼吸振动频率随链或管长的增加而频率移动相反,即分别发生红移和蓝移,但随着链和管趋于无限长频率趋近稳定值;同时,随着直径的增大,振动频率减小。在小尺寸时,由于结构内部的应力影响使频率移动更明显,进一步对键长和键角分析说明,结构应力的作用使频率随尺寸变化发生移动。     

二氧化硅准一维纳米团簇的弯曲振动

运用密度泛函理论,计算了4～24个单元的准一维单链、双链、三元环、四元环管状构型SiO₂团簇的振动频率。在不同构型的红外光谱中,侧重研究了频率小于300cm^-1的低频处的弯曲振动。该类振动的频率和强度有很强的尺寸效应,频率与团簇长度呈指数关系,并且指数随直径变化表明它依赖于团簇的几何维数。实验发现此类红外振动只存在于长度0.8nm～5.5nm的SiO₂团簇中,揭示出实验和理论上可用于表征特定长度范围的纳米团簇。     

洛伦兹光束的传输特性研究

运用光束传输的二阶矩理论,对洛伦兹光束的传输特性进行了研究,给出了束腰、横向发散角和光束传输因子的表达式.横向束腰仅取决于对应方向上的光束参数;而横向发散角和光束传输因子却取决于两横向上的光束参数.给出了光束传输因子随两横向光束参数的变化关系曲线.结果表明,两横向上的光束传输因子随两光束参数的变化规律是不相同的,而整体的光束传输因子随两光束参数的变化是前两者的综合体现;但在傍轴情形下,光束传输因子趋向于常数141,因此,对于相同束腰,其傍轴条件下的发散性为对应高斯光束的141倍.所以,洛伦兹光束适合用于描述某些发散程度较大的激光光源. 关键词： 洛伦兹光束 光束传输 二阶矩     

多孔硅量子点的受激辐射：陷阱态的作用

 经激光辐照和高温退火后能够在硅基上生成氧化多孔硅结构。用514 nm的激光泵浦,观测到该多孔硅的受激辐射。当激励强度超过阈值时,在650～750 nm区域有很强的受激发光峰。这些受激发光峰的半高宽小于0.5 nm。激光辐照和高温退火后,在样品上能形成某些特殊的氧化结构。在傅里叶红外光谱分析中,显示有硅氧双键或硅氧桥键在硅表面形成。计算结果表明：当硅氧双键或硅氧桥键形成时,电子的陷阱态出现在纳晶硅的带隙中。价带顶和陷阱态之间的粒子数反转是解释这种受激辐射的关键。     

Linear ion trap imperfection and the compensation of excess micromotion

Quantum computing requires ultracold ions in a ground vibrational state,which is achieved by sideband cooling.We report our recent efforts towards the Lamb-Dicke regime which is a prerequisite of sideband cooling.We first analyse the possible imperfection in our linear ion trap setup and then demonstrate how to suppress the imperfection by compensating the excess micromotion of the ions.The ions,after the micromotion compensation,are estimated to be very close to the Doppler-cooling limit.     

高能质子在散裂靶中的能量沉积计算与实验验证

高能质子在散裂靶中的能量沉积是散裂靶中子学研究的重要内容之一,准确掌握高能质子在散裂靶中引起的能量沉积分布与瞬态变化是开展散裂靶热工流体设计的重要前提.本文采用MCNPX,PHITS与FLUKA三种蒙特卡罗模拟程序,计算并比较了高能质子入射重金属铅靶、钨靶的能量沉积分布及不同粒子对总能量沉积的占比贡献;针对高能质子入射...     

A transient process observation method based on the non-homogeneous Poisson process model

The current-mode-counting method is a new approach to observing transient processes,especially in transient nuclear fusion,based on the non-homogeneous Poisson process(NHPP)model.In this paper,a new measurement process model of the pulsed radiation field produced by transient nuclear fusion is built based on the NHPP.A simulated measurement is performed using the model,and the current signal from the detector is obtained by simulation based on Poisson process thinning.The neutron time spectrum is reconstructed and is in good agreement with the theoretical value,with its maximum error of a characteristic parameter less than 2.3%.Verification experiments were carried out on a CPNG-6 device at the China Institute of Atomic Energy,with a detection system with a nanosecond response time.The experimental charge amplitude spectra are in good agreement with those obtained by the traditional counting mode,and the characteristic parameters of the time spectrum are in good agreement with the theoretical values.This shows that the current-mode-counting method is effective for the observation of transient nuclear fusion processes.