期刊文献+
共找到3,746篇文章
< 1 2 188 >
每页显示 20 50 100
基于GMM-HMMs与Viterbi回溯的连续手势肌电信号预测与识别
1
作者 杨进兴 刘帅 李俊 《南京信息工程大学学报》 北大核心 2026年第1期11-17,共7页
针对基于表面肌电信号(sEMG)的连续手势识别任务中,存在实时性较差和预测能力不足的问题,提出一种基于GMM-HMMs(高斯混合-隐马尔可夫模型)和Viterbi回溯的连续手势动作识别方法.采用滑动窗口对8通道肌电信号进行分窗,通过GMM-HMMs建立... 针对基于表面肌电信号(sEMG)的连续手势识别任务中,存在实时性较差和预测能力不足的问题,提出一种基于GMM-HMMs(高斯混合-隐马尔可夫模型)和Viterbi回溯的连续手势动作识别方法.采用滑动窗口对8通道肌电信号进行分窗,通过GMM-HMMs建立手势的空闲、上升、稳定和下降4个动作状态,提出改进的Viterbi滑动窗口边缘化策略,建立滑动窗口长期约束,实现连续手势动作状态预测.最终引入最大似然法动态阈值模型以区分手势类别.在由8位实验者完成的包含4种手势的12个连续两手势动作任务中,该方法的平均识别率为98.1%,预测时间为71 ms,明显优于LSTM模型(94.2%,309 ms)和GRU模型(93.8%,300 ms). 展开更多
关键词 模式识别 连续手势 gmm-HMMs Viterbi回溯 表面肌电信号
在线阅读 下载PDF
数字经济对省际绿色水资源利用效率的影响研究——基于系统GMM模型的检验 被引量:2
2
作者 张捷 赵乡阳 +1 位作者 庞庆华 张陈俊 《资源与产业》 2025年第1期11-21,共11页
为了探究数字经济与绿色水资源利用效率之间的内在联系,选取2012—2022年全国31个省份(香港地区、澳门地区和台湾地区除外)的面板数据进行研究,采用包含非期望产出的SBM-DEA模型计算各省份的绿色水资源利用效率,用主成分分析法测算省际... 为了探究数字经济与绿色水资源利用效率之间的内在联系,选取2012—2022年全国31个省份(香港地区、澳门地区和台湾地区除外)的面板数据进行研究,采用包含非期望产出的SBM-DEA模型计算各省份的绿色水资源利用效率,用主成分分析法测算省际数字经济发展水平,运用系统GMM模型研究二者之间的因果关系,实证分析数字经济对绿色水资源利用效率的影响。研究表明:1)从全国范围来看,数字经济对绿色水资源利用效率具有显著的正向促进作用,经济发展水平、财政集中度和市场化水平对提高绿色水资源利用效率有一定作用,而制造业发展水平则在一定程度上阻碍其提高。2)稳健性检验结果表明,采用解释变量再度量、缩尾处理、剔除直辖市、缩短样本年限等4种方法分别进行稳健性检验后,核心解释变量的显著性水平和相关系数正负性均没有变化,数字经济促进绿色水资源利用效率提高的结论存在稳健性。3)区域异质性检验结果表明,数字经济的影响效果在东中西部地区存在显著差异,相对于中西部地区,数字经济对东部地区的绿色水资源利用效率提高存在更为明显的正向推动作用。4)中介效应检验结果表明,三大产业的结构水平变化对绿色水资源利用效率提高有着不同程度的影响,其中第一、第二产业结构升级对数字经济促进绿色水资源利用效率提高起到中介作用。基于上述研究结论,提出如下对策建议:保障产业结构安全稳定,采取差异化的数字经济发展战略;加强数字经济基础设施建设,强化专业性人才队伍建设;加大数字经济创新研发投入,采取综合性措施提档升级。研究结论对推动数字经济发展、提高绿色水资源利用效率、促进我国经济高质量发展具有重要意义。 展开更多
关键词 数字经济 绿色水资源利用效率 系统gmm模型 数据包络分析 SBM-DEA模型
原文传递
GMM聚类与高密度电阻法的道路塌陷隐患探测研究
3
作者 张艳辉 张雨燕 +2 位作者 胡宇佳 罗志彬 赵维刚 《中国安全科学学报》 北大核心 2025年第10期115-123,共9页
为解决道路塌陷隐患探测中高密度电阻法分辨率不足和异常识别精度有限的问题,开展基于高密度电阻法的道路塌陷隐患探测分辨率测试和基于高斯混合模型(GMM)聚类的异常识别方法研究。文中正演采用有限差分法,反演采用高斯-牛顿法,通过数... 为解决道路塌陷隐患探测中高密度电阻法分辨率不足和异常识别精度有限的问题,开展基于高密度电阻法的道路塌陷隐患探测分辨率测试和基于高斯混合模型(GMM)聚类的异常识别方法研究。文中正演采用有限差分法,反演采用高斯-牛顿法,通过数值模拟测试不同电极间距的配置对探测分辨率的影响;结合管网漏损诱发道路塌陷的背景,设计地下病害在不同发展阶段的地电模型,采用GMM聚类分析方法优化高密度电阻法的反演结果。结果表明:调整电极间距和测量参数能显著提高探测分辨率,在4.5 m深度下,缩小电极间距能够有效刻画1 m尺度的地下病害体的位置和形态,且0.5 m电极间距能够兼顾探测精度与计算效率,即探测目标异常体尺度的1/2左右。对于同样埋深的异常体,低阻病害的电阻率值恢复效果优于高阻病害,为不同病害目标探测提供参数优化依据。通过管网漏损诱发的地下空洞模型测试,揭示了高密度电阻法在探测不同阶段漏水病害的可行性,而基于GMM的聚类分析进一步提高了异常区域的识别精度。 展开更多
关键词 高斯混合模型(gmm)聚类 高密度电阻法 道路塌陷 隐患探测 分辨率
原文传递
基于GMM的GB-InSAR图像PS点选择方法
4
作者 田卫明 王龙跃 +1 位作者 高嵩 邓云开 《电子学报》 北大核心 2025年第4期1153-1163,共11页
永久散射体(Permanent Scatterer,PS)点选择是地基干涉合成孔径雷达(Ground-Based Interferometric Synthetic Aperture Radar,GB-InSAR)处理中的关键步骤.现有的PS点选择方法依赖于幅相稳定性或像元之间的高相干性筛选PS点,其中幅相稳... 永久散射体(Permanent Scatterer,PS)点选择是地基干涉合成孔径雷达(Ground-Based Interferometric Synthetic Aperture Radar,GB-InSAR)处理中的关键步骤.现有的PS点选择方法依赖于幅相稳定性或像元之间的高相干性筛选PS点,其中幅相稳定性对相位波动敏感,在一些情况下不能很好地表征PS点的相位误差,而基于高相干性的方法基于局部窗口,容易造成误检.针对上述问题,本文分析了GB-InSAR图像中PS点与非PS点的干涉相位在分布特征上的差异,并基于此提出了一种基于高斯混合模型(Gaussian Mixture Model,GMM)的PS点选择方法.首先在保证质量的前提下,选择足够数量的PS点作为先验参考信息,然后使用GMM拟合参考PS点干涉相位的概率分布,最后依靠全图像元的干涉相位序列与GMM的匹配程度区分PS点与非PS点.实测数据表明,与基于幅相稳定性的传统方法相比,在获得的PS点数量接近的情况下,本文方法获取的PS点的相关性更强,干涉相位序列聚合程度更高,且残差点数量更少. 展开更多
关键词 永久散射体(PS) 地基干涉合成孔径雷达(GB-InSAR) 高斯混合模型(gmm)
在线阅读 下载PDF
基于小波变换和GMM的嗓音特征提取研究
5
作者 魏娜 《自动化与仪器仪表》 2025年第7期24-28,共5页
为了对歌唱嗓音进行特征提取和识别,通过小波变换来获取去噪后的信息,并凭借小波包分解来对梅尔频率倒谱系数发音特征进行改进,以提升歌唱嗓音的识别率。针对歌唱嗓音的识别,采用了改进后的特征和高斯混合通用背景模型。经过测试发现,... 为了对歌唱嗓音进行特征提取和识别,通过小波变换来获取去噪后的信息,并凭借小波包分解来对梅尔频率倒谱系数发音特征进行改进,以提升歌唱嗓音的识别率。针对歌唱嗓音的识别,采用了改进后的特征和高斯混合通用背景模型。经过测试发现,小波包分解对应的识别率最大值为95.57%,改进特征在测试集上的等错误率最小值为0.101%,所设计模型的准确率最大值为98.42%。所设计特征改进方法和模型具有良好的性能,能够对嗓音进行特征提取和识别。 展开更多
关键词 小波变换 gmm-UBM 嗓音 特征 提取 模型
原文传递
基于轨迹多特征凝聚层次聚类和GMM的信号交叉口右转车辆受扰轨迹识别
6
作者 梁国华 张玉洁 +3 位作者 陈亦新 翁康怡 孙润峰 孟霄阳 《长安大学学报(自然科学版)》 北大核心 2025年第2期136-153,共18页
为提升右转车辆驾驶人对信号灯时序、行人或非机动车等干扰因素的应对能力,以实际信号交叉口右转车辆轨迹数据为基础,对右转车辆进行主方向和Hausdorff距离的轨迹多特征凝聚层次聚类,获取不同的右转车辆轨迹运行模式,对轨迹受扰特征变... 为提升右转车辆驾驶人对信号灯时序、行人或非机动车等干扰因素的应对能力,以实际信号交叉口右转车辆轨迹数据为基础,对右转车辆进行主方向和Hausdorff距离的轨迹多特征凝聚层次聚类,获取不同的右转车辆轨迹运行模式,对轨迹受扰特征变化进行统计分析,确定轨迹速度和转向角特征受扰阈值,结合受扰阈值和聚类结果采用高斯混合模型(GMM)对受扰异常轨迹进行识别。研究结果表明:采用基于轨迹主方向和Hausdorff距离的改进凝聚层次聚类算法能够有效刻画不同类别轨迹运动模式,可以将轨迹速度和转向角作为判别轨迹是否受扰的标准,并确定速度受扰阈值为均值减2倍标准差,转向角受扰阈值为均值加减2倍标准差;轨迹多特征凝聚层次聚类和高斯混合模型能有效筛选出右转受扰轨迹,识别精度为99.95%。由此可见,提出方法可以准确判别异常轨迹,可为交叉口内右转车辆未来驾驶风险判断以及驾驶辅助系统设计提供理论依据,为交叉口内车辆安全运行提供预警。 展开更多
关键词 交通工程 车辆轨迹 轨迹多特征凝聚层次聚类 高斯混合模型 异常识别
原文传递
一种基于DTW-DP-GMM的工业机器人轨迹学习策略 被引量:3
7
作者 肖洒 陈旭阳 +1 位作者 叶锦华 吴海彬 《天津大学学报(自然科学与工程技术版)》 EI CAS 北大核心 2025年第1期68-80,共13页
针对机器人示教编程过程中使用高斯混合模型(GMM)规划运动轨迹时存在的高斯分布个数难以选择、复现轨迹精度较低等问题,提出了一种复合的机器人运动轨迹学习策略.该策略包含动态时间规整(DTW)算法、高斯混合模型与道格拉斯-普克(DP)算法... 针对机器人示教编程过程中使用高斯混合模型(GMM)规划运动轨迹时存在的高斯分布个数难以选择、复现轨迹精度较低等问题,提出了一种复合的机器人运动轨迹学习策略.该策略包含动态时间规整(DTW)算法、高斯混合模型与道格拉斯-普克(DP)算法.首先,针对示教过程中采集的多条轨迹在时间长度上存在差异的问题,采用DTW算法来统一示教轨迹在时域上的变化.其次,使用GMM算法对示教轨迹的特征进行提取,并利用高斯混合回归(GMR)算法将其重构为复现轨迹.在这个过程中采用DP算法来预估GMM算法的关键参数高斯分布的数量,与传统方法相比,能够简单直观地得到相对准确的参数值.利用DP算法对复现轨迹的数据点进行稀疏化并优化,不仅确保了机器人最终运动轨迹的精度,而且大幅减少了最终轨迹数据点的数量.最后,进行了不同形状的模拟焊接轨迹学习规划实验.结果表明:经由DTW对齐后的示教轨迹具有更加明显的运动特征,经过GMM-GMR学习输出的复现轨迹具有良好的表征结果;在使用GMM-GMR算法学习示教轨迹的过程中,采用DP算法可以有效预估高斯分布个数;经过DP算法稀疏化并优化的最终轨迹的平均位置误差均在0.500 mm以内,其最大误差可以控制在0.800 mm以内,可以满足焊接轨迹规划的精度要求,验证了该策略的有效性和优越性. 展开更多
关键词 工业机器人 示教编程 高斯混合模型 道格拉斯-普克算法 动态时间规整 轨迹复现
在线阅读 下载PDF
Ascorbic acid-induced porous iodide layer for a high-purity two-step solution-processed tin-lead mixed perovskite photodetector 被引量:2
8
作者 Liansong Liu Fengren Cao +3 位作者 Liukang Bian Meng Wang Haoxuan Sun Liang Li 《Journal of Materials Science & Technology》 2025年第7期227-232,共6页
Tin(Sn)-lead(Pb)mixed halide perovskites have attracted widespread interest due to their wider re-sponse wavelength and lower toxicity than lead halide perovskites,Among the preparation methods,the two-step method mor... Tin(Sn)-lead(Pb)mixed halide perovskites have attracted widespread interest due to their wider re-sponse wavelength and lower toxicity than lead halide perovskites,Among the preparation methods,the two-step method more easily controls the crystallization rate and is suitable for preparing large-area per-ovskite devices.However,the residual low-conductivity iodide layer in the two-step method can affect carrier transport and device stability,and the different crystallization rates of Sn-and Pb-based per-ovskites may result in poor film quality.Therefore,Sn-Pb mixed perovskites are mainly prepared by a one-step method.Herein,a MAPb_(0.5)Sn_(0.5)I_(3)-based self-powered photodetector without a hole transport layer is fabricated by a two-step method.By adjusting the concentration of the ascorbic acid(AA)addi-tive,the final perovskite film exhibited a pure phase without residues,and the optimal device exhibited a high responsivity(0.276 A W^(-1)),large specific detectivity(2.38×10^(12) Jones),and enhanced stability.This enhancement is mainly attributed to the inhibition of Sn2+oxidation,the control of crystal growth,and the sufficient reaction between organic ammonium salts and bottom halides due to the AA-induced pore structure. 展开更多
关键词 PEROVSKITE SN-PB PHOTODETECTOR two-step solution method
原文传递
基于BERT-GMM的空地协同特情处置关键信息提取
9
作者 刘一海 杨昌其 《舰船电子工程》 2025年第6期114-119,共6页
在智慧民航建设的蓝图中,“强安全、强效率、强智慧、强协同”的四强空管是重要组成部分。其中,飞行机组和空中交通管制班组之间良好的空地协同是成功处置民航特情的重要保证。现有的飞行训练和管制培训中对特情处置的培训普遍独立进行... 在智慧民航建设的蓝图中,“强安全、强效率、强智慧、强协同”的四强空管是重要组成部分。其中,飞行机组和空中交通管制班组之间良好的空地协同是成功处置民航特情的重要保证。现有的飞行训练和管制培训中对特情处置的培训普遍独立进行,机组资源管理的训练对象也只是飞行机组与客舱乘务组,并未涉及到空中交通管制班组。针对以上问题,论文从特情处置空地协同的角度出发,在过往民航不安全事件报告数据的基础上,提出一种BERT-GMM融合模型,用于不安全事件文本关键信息提取。通过对比实验表明,该模型在民航不安全事件文本信息提取的ROC-AUC值为0.95,比基于ELMo的GMM融合模型F1值提高了5.4%。 展开更多
关键词 自然语言处理 信息提取 BERT gmm 民航特情
在线阅读 下载PDF
基于惩罚GMM估计的部分线性可加空间自回归模型的变量选择
10
作者 牛潇雪 赵培信 《安庆师范大学学报(自然科学版)》 2025年第2期18-23,共6页
基于惩罚广义矩估计(generalized method of moments,GMM)的正交投影估计方法,文章针对一类部分线性可加空间自回归模型提出了一种变量选择方法。该方法既能在参数分量中选取重要的协变量,又能识别空间效应的显著性。在一定条件下,文章... 基于惩罚广义矩估计(generalized method of moments,GMM)的正交投影估计方法,文章针对一类部分线性可加空间自回归模型提出了一种变量选择方法。该方法既能在参数分量中选取重要的协变量,又能识别空间效应的显著性。在一定条件下,文章亦证明了该方法的相合性,并得到了参数估计量的收敛速度,最后通过蒙特卡罗模拟研究证实了该方法在有限样本下的可行性。 展开更多
关键词 部分线性可加空间自回归模型 变量选择 正交投影 gmm估计
在线阅读 下载PDF
Growth pattern of MAO coating under constant voltage–current two-step power mode 被引量:1
11
作者 Shu-fan Zhou Liang-yu Chen +5 位作者 Wei-gang Lv Jun-jie Gu Fei Ye Dubovyy Oleksandr Sheng Lu Ze-xin Wang 《Journal of Iron and Steel Research International》 2025年第5期1245-1262,共18页
A custom micro-arc oxidation(MAO)apparatus is employed to produce coatings under optimized constant voltage–current two-step power supply mode.Various analytical techniques,including scanning electron microscopy,conf... A custom micro-arc oxidation(MAO)apparatus is employed to produce coatings under optimized constant voltage–current two-step power supply mode.Various analytical techniques,including scanning electron microscopy,confocal laser microscopy,X-ray diffraction,X-ray photoelectron spectroscopy,transmission electron microscopy,and electrochemical analysis,are employed to characterize MAO coatings at different stages of preparation.MAO has MgO,hydroxyapatite,Ca_(3)(PO_(4))_(2),and Mg2SiO4 phases.Its microstructure of the coating is characterized by"multiple breakdowns,pores within pores",and"repaired blind pores".The porosity and the uniformity of MAO coating first declines in the constant voltage mode,then augments while the discharge phenomenon takes place,and finally decreases in the repair stage.These analyses reveal a four-stage growth pattern for MAO coatings:anodic oxidation stage,micro-arc oxidation stage,breakdown stage,and repairing stage.During anodic oxidation and MAO stages,inward growth prevails,while the breakdown stage sees outward and accelerated growth.Simultaneous inward and outward growth in the repair stage results in a denser,more uniform coating with increased thickness and improved corrosion resistance. 展开更多
关键词 Micro-arc oxidation Biomedical material Corrosion resistance Growth pattern two-step power mode
原文传递
Three-dimensional stability of two-step slope with crack considering temperature effect on unsaturated soil 被引量:1
12
作者 SHAN Jun-tao WU Yi-min YANG Xiao-li 《Journal of Central South University》 2025年第3期1060-1079,共20页
In existing studies, most slope stability analyses concentrate on conditions with constant temperature, assuming the slope is intact, and employ the Mohr-Coulomb (M-C) failure criterion for saturated soil to character... In existing studies, most slope stability analyses concentrate on conditions with constant temperature, assuming the slope is intact, and employ the Mohr-Coulomb (M-C) failure criterion for saturated soil to characterize the strength of the backfill. However, the actual working temperature of slopes varies, and natural phenomena such as rainfall and groundwater infiltration commonly result in unsaturated soil conditions, with cracks typically present in cohesive slopes. This study introduces a novel approach for assessing the stability of unsaturated soil stepped slopes under varying temperatures, incorporating the effects of open and vertical cracks. Utilizing the kinematic approach and gravity increase method, we developed a three-dimensional (3D) rotational wedge failure mechanism to simulate slope collapse, enhancing the traditional two-dimensional analyses. We integrated temperature-dependent functions and nonlinear shear strength equations to evaluate the impact of temperature on four typical unsaturated soil types. A particle swarm optimization algorithm was employed to calculate the safety factor, ensuring our method’s accuracy by comparing it with existing studies. The results indicate that considering 3D effects yields a higher safety factor, while cracks reduce slope stability. Each unsaturated soil exhibits a distinctive temperature response curve, highlighting the importance of understanding soil types in the design phase. 展开更多
关键词 3D two-step slope cracks temperature effects UNSATURATION limit analysis
在线阅读 下载PDF
基于BKA-GMM算法的需求侧高价值用户划分及筛选方法
13
作者 余洋 孙梓朔 +3 位作者 王中晶 李君卫 庞淇文 樊蕊 《电力自动化设备》 北大核心 2025年第9期165-173,共9页
针对调峰场景中传统用户分类方法评价标准单一导致的聚类效果不佳、大数据处理效率与精度失衡、划分结果可解释性较低等问题,提出基于黑鸢算法(BKA)优化高斯混合模型(GMM)算法的需求侧高价值用户划分与筛选方法。为解决用户分类不明确... 针对调峰场景中传统用户分类方法评价标准单一导致的聚类效果不佳、大数据处理效率与精度失衡、划分结果可解释性较低等问题,提出基于黑鸢算法(BKA)优化高斯混合模型(GMM)算法的需求侧高价值用户划分与筛选方法。为解决用户分类不明确和系统匹配不良的问题,结合用户心理学与峰、谷分时电价的负荷转移率模型,引入改进S形函数的用户负荷削减率,构建用户激励-潜力模型;针对传统聚类算法在大规模用户场景下精度与效率不足的问题,提出BKA-GMM算法,预先确定最佳聚类数和正则化参数,结合基于峰、谷时段调度潜力的划分依据,实现用户的初步分类;为提高分类结果的解释性,根据潜力系数及3个用户群特征制定筛选条件,精确识别出削减型、灵活型和复合型的高价值用户。仿真结果表明,将所建模型作为划分条件,显著提升了用户划分效率和参与实际响应的概率,所提筛选方法不仅加快了划分速度,提高了准确性,而且清晰地筛选出了不同类型的高价值用户。 展开更多
关键词 调度潜力 高价值用户 BKA优化算法 gmm聚类分析 用户划分 需求响应
在线阅读 下载PDF
Improving the fracture strain of graphite materials by in-situ porosity introduction by two-step sintering
14
作者 GU Shi-jia CHEN Han-lin +3 位作者 WANG Jun-zhuo LU Xiao-fang WANG Lian-jun JIANG Wan 《新型炭材料(中英文)》 北大核心 2025年第3期703-716,共14页
High-performance graphite materials have important roles in aerospace and nuclear reactor technologies because of their outstanding chemical stability and high-temperature performance.Their traditional production meth... High-performance graphite materials have important roles in aerospace and nuclear reactor technologies because of their outstanding chemical stability and high-temperature performance.Their traditional production method relies on repeated impregnation-carbonization and graphitization,and is plagued by lengthy preparation cycles and high energy consumption.Phase transition-assisted self-pressurized selfsintering technology can rapidly produce high-strength graphite materials,but the fracture strain of the graphite materials produced is poor.To solve this problem,this study used a two-step sintering method to uniformly introduce micro-nano pores into natural graphite-based bulk graphite,achieving improved fracture strain of the samples without reducing their density and mechanical properties.Using natural graphite powder,micron-diamond,and nano-diamond as raw materials,and by precisely controlling the staged pressure release process,the degree of diamond phase transition expansion was effectively regulated.The strain-to-failure of the graphite samples reached 1.2%,a 35%increase compared to samples produced by fullpressure sintering.Meanwhile,their flexural strength exceeded 110 MPa,and their density was over 1.9 g/cm^(3).The process therefore produced both a high strength and a high fracture strain.The interface evolution and toughening mechanism during the two-step sintering process were investigated.It is believed that the micro-nano pores formed have two roles:as stress concentrators they induce yielding by shear and as multi-crack propagation paths they significantly lengthen the crack propagation path.The two-step sintering phase transition strategy introduces pores and provides a new approach for increasing the fracture strain of brittle materials. 展开更多
关键词 High-performance graphite Phase transition control two-step sintering process Fracture strain IN-SITU
在线阅读 下载PDF
多阶GMM-Res Net融合在语音伪造检测中的研究
15
作者 曹明明 雷震春 +1 位作者 杨印根 周勇 《信息安全学报》 2025年第2期116-126,共11页
近年来,自动说话人识别技术取得了显著进步,但同时也容易受到合成或转换语音的伪造攻击,语音伪造检测系统致力于解决这一问题。本文根据不同阶数GMM中高斯分量之间的相关性和ResNet模型中不同层次残差块输出的特征信息,提出了一种多阶GM... 近年来,自动说话人识别技术取得了显著进步,但同时也容易受到合成或转换语音的伪造攻击,语音伪造检测系统致力于解决这一问题。本文根据不同阶数GMM中高斯分量之间的相关性和ResNet模型中不同层次残差块输出的特征信息,提出了一种多阶GMM-ResNet融合模型进行语音伪造检测。该模型主要包含两部分:多阶对数高斯概率(Log Gaussian Probability, LGP)特征融合和多尺度特征聚合ResNet(Multi-Scale Feature Aggregation ResNet, MFA-ResNet)。GMM描述了语音特征在其空间的分布情况,不同阶数的GMM则具有不同描述能力来形成对特征分布的平滑近似。此外,根据不同阶数GMM计算出来的LGP特征也就在不同阶上捕获语音信息。多阶LGP特征融合将基于不同阶数的GMM得到的三种不同阶LGP特征进行加权融合,从而促进不同阶LGP特征之间的信息交换。另一方面,神经网络模型中第一层或中间层获得的特征信息对于分类任务也是非常有用的。基于这一经验, MFA-ResNet模块通过对每个ResNet块输出的特征进行聚合,充分融合网络内不同层级的特征信息,从而提高网络的特征提取能力。在ASVspoof 2019逻辑访问场景下, LFCC+多阶GMM-ResNet融合系统的min t-DCF和EER分别为0.0353和1.16%,比基线系统LFCC+GMM分别相对降低了83.3%和85.7%。在ASVspoof2021逻辑访问场景下,LFCC+多阶GMM-ResNet融合系统的min t-DCF和EER分别为0.2459和2.50%,比基线系统LFCC+GMM分别相对降低了57.3%和87.1%,比基线系统LFCC+LCNN分别相对降低了28.6%和73.0%。与目前最先进模型相比,本文模型也非常具有竞争力。 展开更多
关键词 多阶gmm-ResNet融合 多阶对数高斯概率特征融合 多尺度特征聚合 语音伪造检测
在线阅读 下载PDF
Microstructure and mechanical properties of ZK61 magnesium alloy thin-walled cylindrical component processed by two-step forging
16
作者 Fang Chai Jianqiang Feng +6 位作者 Xinghui Han Wuhao Zhuang Yizhe Chen Zhili Hu Xuan Hu Fangyan Zheng Lin Hua 《Journal of Magnesium and Alloys》 2025年第5期2416-2432,共17页
Magnesium alloy thin-walled cylindrical components with the advantages of high specific stiffness and strength present broad prospect for the lightweight of aerospace components.However,poor formability resulting from... Magnesium alloy thin-walled cylindrical components with the advantages of high specific stiffness and strength present broad prospect for the lightweight of aerospace components.However,poor formability resulting from the hexagonal close-packed crystal structure in magnesium alloy puts forwards a great challenge for thin-walled cylindrical components fabrication,especially for extreme structure with the thicknesschanging web and the high thin-wall.In this research,an ZK61 magnesium alloy thin-walled cylindrical component was successfully fabricated by two-step forging,i.e.,the pre-forging and final-forging is mainly used for wed and thin-wall formation,respectively.Microstructure and mechanical properties at the core,middle and margin of the web and the thin-wall of the pre-forged and final-forged components are studied in detail.Due to the large strain-effectiveness and metal flow along the radial direction(RD),the grains of the web are all elongated along RD for the pre-forged component,where an increasingly elongated trend is found from the core to the margin of the wed.A relatively low recrystallized degree occurs during pre-forging,and the web at different positions are all with prismatic and pyramid textures.During finalforging,the microstructures of the web and the thin-wall are almost equiaxed due to the remarkable occurrence of dynamic recrystallization.Similarity,except for few basal texture of the thin-wall,only prismatic and pyramid textures are found for the final-forged component.Compared with the initial billet,an obviously improved mechanical isotropy is achieved during pre-forging,which is well-maintained during final-forging. 展开更多
关键词 Magnesium alloy Thin-walled cylindrical component two-step forging Microstructure Mechanical properties
在线阅读 下载PDF
Novel adaptive IMEX two-step Runge-Kutta temporal discretization methods for unsteady flows
17
作者 Xueyu QIN Jian YU +2 位作者 Xin ZHANG Zhenhua JIANG Chao YAN 《Chinese Journal of Aeronautics》 2025年第8期142-153,共12页
Efficient and accurate simulation of unsteady flow presents a significant challenge that needs to be overcome in computational fluid dynamics.Temporal discretization method plays a crucial role in the simulation of un... Efficient and accurate simulation of unsteady flow presents a significant challenge that needs to be overcome in computational fluid dynamics.Temporal discretization method plays a crucial role in the simulation of unsteady flows.To enhance computational efficiency,we propose the Implicit-Explicit Two-Step Runge-Kutta(IMEX-TSRK)time-stepping discretization methods for unsteady flows,and develop a novel adaptive algorithm that correctly partitions spatial regions to apply implicit or explicit methods.The novel adaptive IMEX-TSRK schemes effectively handle the numerical stiffness of the small grid size and improve computational efficiency.Compared to implicit and explicit Runge-Kutta(RK)schemes,the IMEX-TSRK methods achieve the same order of accuracy with fewer first derivative calculations.Numerical case tests demonstrate that the IMEX-TSRK methods maintain numerical stability while enhancing computational efficiency.Specifically,in high Reynolds number flows,the computational efficiency of the IMEX-TSRK methods surpasses that of explicit RK schemes by more than one order of magnitude,and that of implicit RK schemes several times over. 展开更多
关键词 Implicit-explicit temporal methods two-step Runge-Kutta methods Adaptive algorithm Unsteady flows Navier-Stokes equations
原文传递
Achieving ultrahigh-specific strength and enhanced GFA in Ti-based bulk metallic glasses via a two-step alloying strategy
18
作者 Heng-Tong Bu Jia-Lun Gu +2 位作者 Yun-Shuai Su Yang Shao Ke-Fu Yao 《Rare Metals》 2025年第3期1932-1942,共11页
Ti-based bulk metallic glasses(BMGs)have attracted increasing attention due to their high specific strength.However,a fundamental conflict exists between the specific strength and glass-forming ability(GFA)of Ti-based... Ti-based bulk metallic glasses(BMGs)have attracted increasing attention due to their high specific strength.However,a fundamental conflict exists between the specific strength and glass-forming ability(GFA)of Ti-based BMGs,restricting their commercial applications significantly.In this study,this challenge was addressed by introducing a two-step alloying strategy to mitigate the remarkable density increment effect associated with heavy alloying elements required for enhancing the GFA.Consequently,through two-step alloying with Al and Fe in sequence,simultaneous enhancements in specific strength and GFA were achieved based on a Ti-Zr-Be ternary metallic glass,resulting in the development of a series of centimeter-sized metallic glasses exhibiting ultrahigh-specific strength.Notably,the newly developed(Ti_(45)Zr_(20)Be_(31)A_(l4))_(94)Fe_(6)alloy established a new record for the specific strength of Ti-based BMGs.Along with a critical diameter(D_(c))of 10 mm,it offers the optimal scheme for balancing the specific strength and GFA of Ti-based BMGs.The present results further brighten the application prospects of Ti-based BMGs as lightweight materials. 展开更多
关键词 Ti-based bulk metallic glasses Specific strength Glass-forming ability two-step alloying strategy
原文传递
交通基础设施对旅游发展的影响——基于皖南地区动态面板GMM模型的实证
19
作者 刘炯 《合肥师范学院学报》 2025年第3期59-62,共4页
交通基础设施是旅游业发展的重要支撑条件。选取皖南地区6市2015-2021年相关数据为样本,构建动态面板GMM模型实证探讨交通基础设施对旅游发展的影响。结果表明,旅游发展的滞后1期、交通基础设施、国内旅游人均花费和产业结构均显著地促... 交通基础设施是旅游业发展的重要支撑条件。选取皖南地区6市2015-2021年相关数据为样本,构建动态面板GMM模型实证探讨交通基础设施对旅游发展的影响。结果表明,旅游发展的滞后1期、交通基础设施、国内旅游人均花费和产业结构均显著地促进了旅游发展,政府支出和城市化皆显著地负向影响旅游发展。强化交通基础设施建设,发挥特色旅游资源优势,加大政府对旅游业的支持力度,有利于促进旅游业快速发展。 展开更多
关键词 交通基础设施 旅游 gmm模型 皖南地区
在线阅读 下载PDF
Ultrafine L10 PtFeZn intermetallics via a two-step annealing process for oxygen reduction reaction:Decoupling alloying and ordering stages
20
作者 Yun-Fei Xia Bo Liu +9 位作者 Zi-Yu Zhang Zi-Gang Zhao Pan Guo Si Lin Bing Liu Yan Wang Yun-Long Zhang Lei Zhao Li-Guang Wang Zhen-Bo Wang 《Chinese Journal of Catalysis》 2025年第11期324-335,共12页
In this paper,we report the design of ultrafine ordered PtFeZn ternary intermetallics uniformly supported on ZIF-8-derived Zn,N-codoped graphitic carbon(ZnNC)via a green aqueous impregnation method followed by a two-s... In this paper,we report the design of ultrafine ordered PtFeZn ternary intermetallics uniformly supported on ZIF-8-derived Zn,N-codoped graphitic carbon(ZnNC)via a green aqueous impregnation method followed by a two-step annealing protocol(H_(2)/Ar,600 and 800℃)to circumvent the sintering issues imposed by conventional thermodynamics.Physical characterizations(X-ray diffraction,high-angle annular dark-field scanning transmission electron microscopy,X-ray absorption spectroscopy)and theoretical calculations reveal that low-temperature annealing at 600℃stabilizes sub-nano disordered PtFe alloys via the strong metal-support interactions(SMSI)between Zn in ZnNC and Pt precursors,while high-temperature treatment at 800℃promotes Zn diffusion from the support into the alloy bulk and simultaneously triggers the disorder-to-order phase transition.The as-prepared ZnNC-15PtFeZn exhibits an initial mass activity of 0.769 mA/μgPt and retains 61.7%of its activity after 30000 cycles of accelerated stress testing(AST).Notably,when used as a cathode catalyst in MEA,ZnNC-15PtFeZn achieves superior power density(2.018 W/cm^(2)under H_(2)-O_(2))at half the Pt loading(0.05 mg/cm^(2))of state-of-the-art commercial Pt/C,highlighting its potential for low-Pt PEMFCs.Density functional theory confirms that Fe enhances ORR activity via ligand effects,while Zn strengthens Pt-Fe/Zn bonding(elevating vacancy formation energies),thereby improving structural stability.This mild,scalable aqueous impregnation strategy offers a general approach for synthesizing multi-component ordered alloys in electrocatalysis. 展开更多
关键词 Oxygen reduction reaction Zn-NC support PtFeZn ternary intermetallic two-step annealing Strong-metal support interaction
在线阅读 下载PDF
上一页 1 2 188 下一页 到第
使用帮助 返回顶部