视频驱动的语义表情基动画方法

为了重用视频内容中的表情信息,提出一种语义表情构造与语义表情参数优化方法.首先从带有噪声的稀疏特征点中定义出人脸表情的语义信息;然后在语义表情空间中优化求解出最优表情参数,以提高人脸动画的真实感.该方法既不需要标定相机参数,也不需要预先建立表演者的3D人脸模型及其表情基,因此除了可用于网络视频的表情重用,也可用于开发实时在线的网络社交等应用.实验结果表明,对于头部摆动的俯仰角和侧角在[?15?,15?]范围内的原始视频,文中方法能够实时合成稳定、逼真的表情动画.     

一种对象代理数据库的跨类查询优化方法

对象代理数据库中跨类查询可以充分发挥对象代理模型的灵活性,为用户提供个性化数据服务,其执行效率十分重要。然而在处理多个跨类属性查询时,现有基于路径表达式的跨类查询实现存在对公共路径节点对象进行重复获取的情形,执行效率较低。针对跨类查询中加快获取终点对象的问题,优化核心思想是减少对路径上节点对象的重复与不必要的遍历,包括两个关键策略:首先是将路径节点整体作为虚拟路径视图统一获取节点对象,避免了多跨类属性查询下公共路径节点的冗余遍历;其次是针对路径复杂过长的跨类查询,依据代价估计策略选择物化查询涉及起点与终点对象,利用缓存减少执行时路径上中间节点的遍历。分别在属性数目与结果集规模两方面进行了对比实验,实验结果表明了优化方法的有效性。     

基于表情传输的交互式照片编辑

目的 生活中照片拍摄时难以捕捉到所有人脸表情最佳的时刻,多次摆拍不仅费时,而且可能会错过某些场景,传统的后期编辑软件不具备针对性,且操作复杂。针对人物照片中部分人脸表情不佳的情况,提出一种基于表情传输的交互式照片编辑算法。方法 首先将包含源人脸的照片与具有目标表情人脸的照片进行特征点检测,通过交互将指定人脸部分选出并将其姿态归正为眼睛在同一水平线上,如果目标人脸与源人脸身份相同,将目标人脸区域根据源人脸的轮廓以及左右半脸分布以扫描线变形得到替换目标;不相同时根据目标人脸特征点分布的几何特征计算源人脸中特征点的新位置,通过基于特征点变化的网格变形得到替换目标,最后利用二次光照与泊松融合将其无缝拼接到源图中。结果 实验表明算法可以对人脸五官清晰且在宽容度内的人物照片进行表情编辑,处理结果只改变了人物的脸部表情,并且无明显拼接痕迹。结论 提出了一种新型针对目标人脸不同身份信息的交互式表情传输模型,该模型可以适应不同的编辑条件与要求,效果出色。     

The future of targeted peptidomics

Targeted MS is becoming increasingly important for sensitive and specific quantitative detection of proteins and respective PTMs. In this article, Ceglarek et al. [Proteomics Clin. Appl. 2013, 7, 794–801] present an LC-MS-based method for simultaneous quantitation of seven apolipoproteins in serum specimens. The assay fulfills many necessities of routine diagnostic applications, namely, low cost, high throughput, and good reproducibility. We anticipate that validation of new biomarkers will speed up with this technology and the palette of laboratory-based diagnostic tools will hopefully be augmented significantly in the near future.     

The contribution of proteomics towards deciphering the enigma of Coxiella burnetii

Coxiella burnetii, the causative agent of Q fever, is an intracellular bacterium and a potential weapon for bioterrorism. The widespread throughout the world, zoonosis is manifested clinically as a self-limited febrile illness, as pneumonia (acute Q fever) or as a chronic illness with endocarditis being its major complication. The recent Netherlands Q fever outbreak has driven the bacterium from a relatively cryptic, underappreciated, “niche” microorganism on the sideline of bacteriology, to one of possibly great impact on public health. Advances in the study of this microorganism proceeded slowly, primarily due to the, until recently, obligatory intracellular nature of the pathogen that in its virulent phase I must be manipulated under biosafety level-3 conditions. Proteomic studies, in particular, have generated a vast amount of information concerning several aspects of the bacterium such as virulence factors, detection/diagnostic and immunogenic biomarkers, inter-/intraspecies variation, resistance to antibiotics, and secreted effector proteins with significant clinical impact. The phenomenon observed following the genomics era, that of generation and accumulation of huge amount of data that ultimately end up unexploited on several databases, begins to emerge in the proteomics field as well. This review will focus on the advances in the field of C. burnetii proteomics through MS, attempting in parallel to utilize some of the proteomics findings by suggesting future directions for the improvement of Q fever diagnosis and therapy.     

Molecular interaction networks in the analyses of sequence variation and proteomics data

Protein–protein interaction networks are typically generated in standard cell lines or model organisms as it is prohibitively difficult to record large interaction datasets from specific tissues or disease models at a reasonable pace. Although the interaction data are of high confidence, they thus do not reflect in vivo relationships as such. A wealth of physiologically relevant protein information, obtained under different conditions and from different systems, is available including information on genetic variation, protein levels, and PTMs. However, these data are difficult to assess comprehensively because the relationships between the entities remain elusive from the measurements. Here, we exemplarily highlight recent studies that gained deeper insight from genetic variation, protein, and PTM measurements using interaction information pointing toward the importance and potential of interaction networks for the interpretation of sequencing and proteomics data.     

综合利用水利工程全寿命周期管理与资产效益发挥

根据全寿命周期管理理论,结合水利工程设计和建设管理实际,论述在综合利用水利工程资产效益发挥中引入全寿命周期管理的必要性,从功能、造价和建管水平三要素,系统分析了全寿命周期管理与综合利用水利工程资产效益发挥的关系,提出了更好发挥水利工程资产效益的措施和建议。     

基于融合信息的癌症相关基因选择方法

基因表达数据的出现,为人类从分子生物学的角度研究和探索癌症的发病机理提供了广阔的前景,利用基因表达数据发现与癌症相关的基因对于癌症的诊断和治疗具有重要的意义。在过去的十几年里,已经有很多种计算方法被成功地用于从基因表达数据中找出与癌症相关的关键基因,然而,不同的方法从不同的角度刻画基因对不同类型样本的区分能力,它们选择出来的关键基因可能不一致,这将给医学解释和应用带来困扰。现提出一种融合的方法,即将基因在不同方面对样本的判别能力结合起来,首先计算每个基因的信息增益、全局判别能力和局部判别能力,再用它们的识别率进行加权,进而计算每个基因的综合判别能力,最后筛选出判别能力最高的基因子集作为关键基因子集。实验结果表明,此方法得到了比采用单独一种评价标准更好的识别效果。     

基于相容关系的基因选择方法

有效的基因选择是对基因表达数据进行分析的重要内容。粗糙集作为一种软计算方法能够保持在数据集分类能力不变的基础上,对属性进行约简。由于基因表达数据的连续性,为了避免运用粗糙集方法所必需的离散化过程带来的信息丢失,将相容粗糙集应用于基因的特征选取,提出了基于相容关系的基因选择方法。首先,通过i检验对基因表达数据进行排列,选择评分靠前的若干基因;然后,通过相容粗糙集对这些基因进一步约简。在两个标准的基因表达数据上进行了实验,结果表明该方法是可行性和有效性的。     

基于邻域粗糙集的加权KNN肿瘤基因表达谱分类算法

肿瘤亚型的准确判别对肿瘤的治疗具有重要意义,对肿瘤的不同亚型进行准确判别是当前生物信息学研究的重要课题．本文首先利用Relief算法排序基因并选出初始的肿瘤信息基因子集,然后利用向基于邻域粗糙集模型的向前属性约减算法FARNeM来计算加权基因集合,最后用加权KNN算法对肿瘤对这些数据进行分析,从而发现有差异的基因表达。实验结果表明了上述方法的可行性和有效性。