FREQUENCY DOMAIN TESTS OF MULTIVARIATE GAUSSIANITY AND LINEARITY

A stationary multivariate time series { X _t } is defined as linear if it can be written in the form X _t = ∑^∞ _{j =−∞} A _j e _{t − j} where A _j are square matrices and e _t are independent and identically distributed random vectors. If the e _t } are normally distributed, then { X _t is a multivariate Gaussian linear process. This paper is concerned with the testing of departures of a vector stationary process from multivariate Gaussianity and linearity using the bispectral approach. First the definition and properties of cumulants of random matrices are used to obtain the expressions for the higher-order cumulant and spectral vectors of a linear vector process as defined above. Then it is shown that linearity of a vector process implies constancy of the modulus square of its normalized higher-order spectra whereas the component of such a vector process does not necessarily have a linear representation. Finally, statistics for the testing of multivariate Gaussianity and linearity are proposed.     

A low-cost real time virtual system for postural stability assessment at home

Background and objective

The degeneration of the balance control system in the elderly and in many pathologies requires measuring the equilibrium conditions very often. In clinical practice, equilibrium control is commonly evaluated by using a force platform (stabilometric platform) in a clinical environment.In this paper, we demonstrate how a simple movement analysis system, based on a 3D video camera and a 3D real time model reconstruction of the human body, can be used to collect information usually recorded by a physical stabilometric platform.

Methods

The algorithm used to reconstruct the human body model as a set of spheres is described and discussed. Moreover, experimental measurements and comparisons with data collected by a physical stabilometric platform are also reported. The measurements were collected on a set of 6 healthy subjects to whom a change in equilibrium condition was stimulated by performing an equilibrium task.

Results

The experimental results showed that more than 95% of data collected by the proposed method were not significantly different from those collected by the classic platform, thus confirming the usefulness of the proposed system.

Conclusions

The proposed virtual balance assessment system can be implemented at low cost (about 500$) and, for this reason, can be considered a home use medical device. On the contrary, astabilometric platform has a cost of about 10,000$ and requires periodical calibration. The proposed system does not require periodical calibration, as is necessary for stabilometric force platforms, and it is easy to use. In future, the proposed system with little integration can be used, besides being an emulator of a stabilometric platform, also to recognize and track, in real time, head, legs, arms and trunk, that is to collect information actually obtained by sophisticated optoelectronic systems.     

ASYMPTOTIC PROPERTIES OF SOME PRELIMINARY ESTIMATORS FOR AUTOREGRESSIVE MOVING AVERAGE TIME SERIES MODELS

Abstract. Some simple preliminary estimators for the coefficients of mixed autoregressive moving average time series models are considered. As the first step the estimators require the fitting of a long autoregression to the data. The first two methods of the paper are non-iterative and generally inefficient. The estimators are Yule-Walker type modifications of the least squares estimators of the coefficients in auxiliary linear regression models derived, respectively, for the coefficients of the long autoregression and for the coefficients of the corresponding long moving average approximation of the model. Both of these estimators are shown to be strongly consistent and their asymptotic distributions are derived. The asymptotic distributions are used in studying the loss in efficiency and in constructing the third estimator of the paper which is an asymptotically efficient two-step estimator. A numerical illustration of the third estimator with real data is given.     

智能化结构分析建模方法

建立正确的模型是进行机械结构分析的关键技术之一。提出了一种基于知识的结构分析建模方法,给出了基于知识结构分析建模方法的总体结构,讨论了知识的获取、分类与表达,提出了推理机制以及智能求解策略,提高了结构有限元分析建模方法的实用性和有效性。最后用实例说明了应用效果。     

过程数据的分段线性表示

在时间序列分段线性表示(PLR)基础上，提出用相对点平均误差度量子序列过程状态变化程度，改进过程数据PLR模型分段算法，克服采用单一误差算法的模型失配问题，更加准确地反应过程状态的变化。     

基于稀疏表示的图像超分辨率重构算法

为提高单幅图像的分辨率,提出一种基于稀疏表示的图像超分辨率重构方法。该方法的核心是联合训练高分辨率和低分辨率字典,然后利用所得字典求解高、低分辨率下图像块共有的稀疏表示系数。与已有的基于稀疏表示的图像超分辨重构算法相比,该算法在求解稀疏表示系数时并未采用拉格朗日乘子将稀疏度和重构误差相结合,而是利用对偶模型求解原始的带约束优化问题。实验表明,与其他图像超分辨率重构方法相比,该方法所需手动调节参数较少,重构效果较好。     

基于稀疏表征的话者识别

近年来,随着信号的稀疏性理论越来越受到人们的关注,稀疏表征分类器也作为一种新型的分类算法被应用到话者识别系统中。该模型的基本思想是:只要超完备字典足够大,任意待测样本都能够用超完备字典进行线性表示。基于信号的稀疏性理论,未知话者的向量系数,即稀疏解可以通过L1范数最小化获取。超完备字典则可视为语音特征向量在高斯混合模型-通用背景模型(GMM-UBM)上进行MAP自适应而得到的大型数据库。采用稀疏表征模型作为话者辨认的分类方法,基于TIMIT语料库的实验结果表明,所采用的话者辨认方法,能够大大提高说话人识别系统的性能。     

视觉特征对比解耦的广义零样本学习

广义零样本学习通常利用在ImageNet上预训练的深度模型来提取相应的视觉特征,然而预训练模型提取到的视觉特征不可避免地包含和语义无关的信息,这将导致语义—视觉对齐的偏差以及对不可见类的负迁移,从而影响分类结果。为解决上述问题,提出了视觉特征对比解耦的广义零样本学习模型(visual feature contrast decoupling for generalized zero-shot learning, VFCD-GZSL),通过解耦出视觉特征中的语义相关表示来降低冗余信息对分类结果的影响。具体来说,首先用条件变分自编码器生成不可见类的视觉特征。然后通过解耦模块将视觉特征解耦语义相关和语义无关的潜层表示,同时添加总相关惩罚和对比损失来鼓励两者间的相互独立,并用语义关系匹配模型衡量其语义一致性,从而指导模型学习语义相关表示。最后使用特征细化模块细化后的特征和语义相关表示联合学习一个广义零样本学习分类器。在四个数据集上的实验均取得较优的结果,证实了所提方法的有效性。     

图表示学习方法研究综述

针对图表示方法的相关解析任务进行了研究,从形式化定义出发,首先以不同核心技术作为分类标准将图表示学习方法划分为五大类,其包括基于降维解析、矩阵分解、随机游走、深度学习和其他表示学习方法。其次通过归纳与对比分析梳理各类技术发展脉络,进而深层次展现各类图表示方法的优劣。随后结合图表示学习的常用数据集、评估方法和应用领域的归纳分析,展开动态性、可扩展性、可解释性和可解析性的四维剖析。最后总结并展望了图表示学习的未来研究趋势与发展方向。     

基于分布的中文词表示研究

针对基于分布的中文词表示构造过程中的参数选择问题进行了系统性的研究。选择了六种参数进行对比实验,在中文语义相似度任务上对不同参数设置下得到的中文词表示的质量进行了评估。实验结果表明,通过选择合适的参数,基于分布的词表示在中文语义相似度任务上能够得到较高的性能,而且,这种高维的词分布表示的质量甚至优于目前流行的基于神经网络（Skip-gram）或矩阵分解（GloVe）得到的低维的词表示。