文章研究了背景为子空间干扰加高斯杂波的距离扩展目标方向检测问题。杂波是均值为零协方差矩阵未知但具有斜对称特性的高斯杂波,目标与干扰分别通过具备斜对称特性的目标子空间和干扰子空间描述。针对方向检测问题,利用上述斜对称性,...文章研究了背景为子空间干扰加高斯杂波的距离扩展目标方向检测问题。杂波是均值为零协方差矩阵未知但具有斜对称特性的高斯杂波,目标与干扰分别通过具备斜对称特性的目标子空间和干扰子空间描述。针对方向检测问题,利用上述斜对称性,根据广义似然比检验(Generalized Likeli-hood Ratio Test,GLRT)准则的一步与两步设计方法,设计了基于GLRT的一步法与两步法的距离扩展目标方向检测器。通过理论推导证明了这2种检测器相对于未知杂波协方差矩阵都具有恒虚警率。对比相同背景下已有检测器,特别是在辅助数据有限的场景下,文章提出的2个检测器表现出了优越的检测性能。展开更多
Responses of a Micro-hotplate based integrated gas sensor array to CO and CH4 were measured with an automated gas sensor calibration system.Combining with the blind source separation（BSS） techniques,the blind separa...Responses of a Micro-hotplate based integrated gas sensor array to CO and CH4 were measured with an automated gas sensor calibration system.Combining with the blind source separation（BSS） techniques,the blind separability in gas mixture analysis was discussed.The widely used BSS approach-Independent Component Analysis（ICA） was adopted to verify the proposed method by analyzing the gas mixtures of CO and CH4.The analysis results demonstrate that BSS was an effective way to extract the information of gas components in mixtures,from which the gas concentrations can be estimated.The average relative quantification errors were 9.37% and 8.11% for CO and CH4,respectively,in the specified concentration ranges.展开更多
非侵入式负荷监测(non-intrusive load monitoring,NILM)技术绿色节能,已成为电力系统负荷监测的发展趋势。集成学习方法可有效提高负荷识别性能,但其基学习器的优化选择和权重设置问题亟待解决。文中以一种典型智能电表对8种小型用电...非侵入式负荷监测(non-intrusive load monitoring,NILM)技术绿色节能,已成为电力系统负荷监测的发展趋势。集成学习方法可有效提高负荷识别性能,但其基学习器的优化选择和权重设置问题亟待解决。文中以一种典型智能电表对8种小型用电设备及其混合负荷的高频实测实验为基础,基于递归特征消除(recursive feature elimination,RFE)法选择最优特征组合,提出结合准确率和多样性权衡的基学习器组合优化方法,并引入香农熵设置投票权重,形成一种新颖的基于香农熵加权投票的集成式NILM识别方法。通过在自测数据集和公开的全球家庭和行业瞬态能量数据集(worldwide household and industry transient energy dataset,WHITED)验证,与常用集成方法比较,该方法识别准确率高、运行时间短且稳定性高。展开更多
文摘文章研究了背景为子空间干扰加高斯杂波的距离扩展目标方向检测问题。杂波是均值为零协方差矩阵未知但具有斜对称特性的高斯杂波,目标与干扰分别通过具备斜对称特性的目标子空间和干扰子空间描述。针对方向检测问题,利用上述斜对称性,根据广义似然比检验(Generalized Likeli-hood Ratio Test,GLRT)准则的一步与两步设计方法,设计了基于GLRT的一步法与两步法的距离扩展目标方向检测器。通过理论推导证明了这2种检测器相对于未知杂波协方差矩阵都具有恒虚警率。对比相同背景下已有检测器,特别是在辅助数据有限的场景下,文章提出的2个检测器表现出了优越的检测性能。
文摘Responses of a Micro-hotplate based integrated gas sensor array to CO and CH4 were measured with an automated gas sensor calibration system.Combining with the blind source separation（BSS） techniques,the blind separability in gas mixture analysis was discussed.The widely used BSS approach-Independent Component Analysis（ICA） was adopted to verify the proposed method by analyzing the gas mixtures of CO and CH4.The analysis results demonstrate that BSS was an effective way to extract the information of gas components in mixtures,from which the gas concentrations can be estimated.The average relative quantification errors were 9.37% and 8.11% for CO and CH4,respectively,in the specified concentration ranges.
文摘非侵入式负荷监测(non-intrusive load monitoring,NILM)技术绿色节能,已成为电力系统负荷监测的发展趋势。集成学习方法可有效提高负荷识别性能,但其基学习器的优化选择和权重设置问题亟待解决。文中以一种典型智能电表对8种小型用电设备及其混合负荷的高频实测实验为基础,基于递归特征消除(recursive feature elimination,RFE)法选择最优特征组合,提出结合准确率和多样性权衡的基学习器组合优化方法,并引入香农熵设置投票权重,形成一种新颖的基于香农熵加权投票的集成式NILM识别方法。通过在自测数据集和公开的全球家庭和行业瞬态能量数据集(worldwide household and industry transient energy dataset,WHITED)验证,与常用集成方法比较,该方法识别准确率高、运行时间短且稳定性高。
