We show that self-assembled vertically aligned gold nanorod （VA-GNRs） superlattices can serve as probes or substrates for ultra-high sensitive detection of various molecules. D-glucose and 2,4,6-trinitrotoluene （TN...We show that self-assembled vertically aligned gold nanorod （VA-GNRs） superlattices can serve as probes or substrates for ultra-high sensitive detection of various molecules. D-glucose and 2,4,6-trinitrotoluene （TNT） have been chosen as model systems due to their very low Raman cross-sections （5.6× 10-30 cm2.molecule-1.sr-1 for D-glucose and 4.9 × 10-31 cm2.molecule-1.sr-1 for TNT） to show that the VA-GNR superlattice assembly offers as low as yoctomole sensitivity. Our experiment on mixed samples of bovine serum albumin （BSA） and D-glucose solutions demonstrate sensitivity for the latter, and the possible extension to real samples. Self-assembled superlattices of VA-GNRs were achieved on a silicon wafer by depositing a drop of solvent containing the GNRs and subsequent solvent evaporation in ambient conditions. An additional advantage of the VA-GNR monolayers is their extremely high reproducible morphology accompanied by ultrahigh sensitivity which will be useful in many fields where a very small amount of analyte is available. Moreover the assembly can be reused a number of times after removing the already present molecules. The method of obtaining VA-GNRs is simple, inexpensive and reproducible. With the help of simulations of monolayers and multilayers it has been shown that superlattices can achieve better sensitivity than monolaver assembly of VA-GNRs.展开更多
目的抑郁症是一种严重的精神类障碍,会显著影响患者的日常生活和工作。目前的抑郁症临床评估方法几乎都依赖于临床访谈或问卷调查,缺少系统有效地挖掘与抑郁症密切相关模式信息的手段。为了有效帮助临床医生诊断患者的抑郁症严重程度,...目的抑郁症是一种严重的精神类障碍,会显著影响患者的日常生活和工作。目前的抑郁症临床评估方法几乎都依赖于临床访谈或问卷调查,缺少系统有效地挖掘与抑郁症密切相关模式信息的手段。为了有效帮助临床医生诊断患者的抑郁症严重程度,情感计算领域涌现出越来越多的方法进行自动化的抑郁症识别。为了有效挖掘和编码人们面部含有的具有鉴别力的情感信息,本文提出了一种基于动态面部特征和稀疏编码的抑郁症自动识别框架。方法在面部特征提取方面,提出了一种新的可以深层次挖掘面部宏观和微观结构信息的动态特征描述符,即中值鲁棒局部二值模式—3D正交平面(median robust local binary patterns from three orthogonal planes,MRELBP-TOP)。由于MRELBP-TOP帧级特征的维度较高,且含有部分冗余信息。为了进一步去除冗余信息和保留关键信息,采用随机映射(random projection,RP)对帧级特征MRELBP-TOP进行降维。此外,为了进一步表征经过降维后的高层模式信息,采用稀疏编码(sparse coding,SC)来抽象紧凑的特征表示。最后,采用支持向量机进行抑郁程度的估计,即预测贝克抑郁分数(the Beck depression inventory-II,BDI-II)。结果在AVEC2013(the continuous audiovisual emotion and depression 2013)和AVEC2014测试集上,抑郁程度估计值与真实值之间的均方根误差(root mean square error,RMSE)分别为9.70和9.22,相比基准算法,识别精度分别提高了29%和15%。实验结果表明,本文方法优于当前大多数基于视频的抑郁症识别方法。结论本文构建了基于面部表情的抑郁症识别框架,实现了抑郁程度的有效估计;提出了帧级特征描述子MRELBP-TOP,有效提高了抑郁症识别的精度。展开更多
文摘We show that self-assembled vertically aligned gold nanorod （VA-GNRs） superlattices can serve as probes or substrates for ultra-high sensitive detection of various molecules. D-glucose and 2,4,6-trinitrotoluene （TNT） have been chosen as model systems due to their very low Raman cross-sections （5.6× 10-30 cm2.molecule-1.sr-1 for D-glucose and 4.9 × 10-31 cm2.molecule-1.sr-1 for TNT） to show that the VA-GNR superlattice assembly offers as low as yoctomole sensitivity. Our experiment on mixed samples of bovine serum albumin （BSA） and D-glucose solutions demonstrate sensitivity for the latter, and the possible extension to real samples. Self-assembled superlattices of VA-GNRs were achieved on a silicon wafer by depositing a drop of solvent containing the GNRs and subsequent solvent evaporation in ambient conditions. An additional advantage of the VA-GNR monolayers is their extremely high reproducible morphology accompanied by ultrahigh sensitivity which will be useful in many fields where a very small amount of analyte is available. Moreover the assembly can be reused a number of times after removing the already present molecules. The method of obtaining VA-GNRs is simple, inexpensive and reproducible. With the help of simulations of monolayers and multilayers it has been shown that superlattices can achieve better sensitivity than monolaver assembly of VA-GNRs.
文摘目的抑郁症是一种严重的精神类障碍,会显著影响患者的日常生活和工作。目前的抑郁症临床评估方法几乎都依赖于临床访谈或问卷调查,缺少系统有效地挖掘与抑郁症密切相关模式信息的手段。为了有效帮助临床医生诊断患者的抑郁症严重程度,情感计算领域涌现出越来越多的方法进行自动化的抑郁症识别。为了有效挖掘和编码人们面部含有的具有鉴别力的情感信息,本文提出了一种基于动态面部特征和稀疏编码的抑郁症自动识别框架。方法在面部特征提取方面,提出了一种新的可以深层次挖掘面部宏观和微观结构信息的动态特征描述符,即中值鲁棒局部二值模式—3D正交平面(median robust local binary patterns from three orthogonal planes,MRELBP-TOP)。由于MRELBP-TOP帧级特征的维度较高,且含有部分冗余信息。为了进一步去除冗余信息和保留关键信息,采用随机映射(random projection,RP)对帧级特征MRELBP-TOP进行降维。此外,为了进一步表征经过降维后的高层模式信息,采用稀疏编码(sparse coding,SC)来抽象紧凑的特征表示。最后,采用支持向量机进行抑郁程度的估计,即预测贝克抑郁分数(the Beck depression inventory-II,BDI-II)。结果在AVEC2013(the continuous audiovisual emotion and depression 2013)和AVEC2014测试集上,抑郁程度估计值与真实值之间的均方根误差(root mean square error,RMSE)分别为9.70和9.22,相比基准算法,识别精度分别提高了29%和15%。实验结果表明,本文方法优于当前大多数基于视频的抑郁症识别方法。结论本文构建了基于面部表情的抑郁症识别框架,实现了抑郁程度的有效估计;提出了帧级特征描述子MRELBP-TOP,有效提高了抑郁症识别的精度。
