基于状态估计的摩擦模糊建模与鲁棒自适应控制

针对一类多自由度机械系统, 研究了基于状态估计的摩擦模糊建模与鲁棒自适应控制问题. 提出用模糊状态估计器估计摩擦模型中的不可测变量, 并用严格正实和李雅普诺夫稳定性理论证明了状态估计误差的一致最终有界性. 运用模糊状态估计结果设计了多变量鲁棒自适应控制器, 其中摩擦模糊模型中的自适应参数是基于李雅普诺夫稳定性理论设计的, 并证明了闭环系统跟踪误差的一致最终有界性. 本文对多自由度质量、弹簧和摩擦阻尼系统进行的仿真, 结果表明所提出的状态估计算法和自适应控制策略是有效的.     

IP报文总长字段统计属性研究

由于网络技术的迅速发展及吉比特以太网的出现,对流量进行全分组测量越来越困难,采用抽样方法并以抽样样本来估计网络流量的总体特性已成为研究重点.一般情况下,总体服从什么分布是未知的,对总体是否服从某个分布作检验,常用的方法是x^2分布检验法.然而,当样本中IP报文长度比较集中时,大量实验结果表明IP报文的总长字段不服从x^2分布,但并不能认为抽样样本的统计分布和总体的统计分布就不相同.本文提出了IP报文总长字段的一种新的检验方法——累计偏差检验法,并在南华大学校园网总出口上进行了抽样实验,结果证明抽样样本的分布和总体的分布相同,同时也证明了这种检验方法是非常有效的.     

常减压装置初馏塔顶产品干点软测量的应用研究

在常减压装置中,影响初馏塔顶石脑油干点的因素很多,反应十分复杂,故难以建立准确的机理模型.针对传统的主元分析法(PCA)或自适应模糊推理系统(ANFIS)建立软测模型中的缺点,本文提出采用主元分析法预处理输入变量,再结合自适应模糊推理系统,进行常减压装置初馏塔顶石脑油干点的软测量模型的改进,能及时测定化工过程的变量,对稳定生产过程,有效控制产品质量具有重要意义.通过MATLAB仿真,表明该改进型方法的软测建模效果较好,建模的训练时间大大节省了,且泛化能力和拟合精度很好.     

基于非局部自相似的Shearlet自适应收缩图像去噪

针对Shearlet收缩去噪引入的Gibbs伪影和"裂痕"现象,提出一种结合非局部自相似的Shearlet自适应收缩图像去噪方法.首先,对噪声图像进行多方向多尺度的Shearlet分解;然后,基于高斯比例混合(GSM)模型的Shearlet系数分布建模,利用贝叶斯最小二乘估计对Shearlet系数进行自适应收缩去噪,重构得到初始去噪图像;最后,利用非局域自相似模型对初始去噪图像进行滤波处理,得到最终的去噪图像.实验结果表明,所提方法在更好地保留边缘特征的同时,有效地去除噪声和收缩去噪引入的Gibbs伪影,该方法获得的峰值信噪比(PSNR)和结构自相似指标(SSIM)比基于非抽样剪切波变换(NSST)的硬阈值去噪方法提高1.41 dB和0.08;比非抽样Shearlet域GSM模型去噪方法提高1.04 dB和0.045;比基于三变量模型的剪切波去噪方法提高0.64 dB和0.025.     

基于总体变化子空间自适应的i-vector说话人识别系统研究

在说话人识别研究中,基于身份认证矢量（identity vector,i-vector） 的子空间建模被证明是目前最前沿最有效的说话人建模技术,其中如何有效准确地估计总体变化子空间矩阵T 成为影响系统性能好坏的关键问题. 本文针对i-vector 技术如何在新的应用环境下进行总体变化子空间矩阵T 的自适应估计问题进行了研究,并提出了两种行之有效的自适应估计算法. 在由美国国家标准技术局（American NationalInstitute of Standard and Technology,NIST） 组织的2008 年说话人识别核心评测数据库以及自行采集的测试数据库上的实验结果显示,不论采用测试集数据本身还是与测试集较匹配的开发集数据,通过本文所提的自适应算法来更新总体变化子空间矩阵均可以使更新后的子空间更有利于新测试数据下的低维子空间描述,在新的测试环境下都更有利于说话人分类. 此外实验结果还表明基于多子空间拼接的子空间自适应方法性能明显优于迭代自适应方法,而且两者的结合可达到最优的识别性能,且此时利用开发集数据进行自适应可以接近其利用测试集数据进行自适应得到的最优性能.     

基于更新样本智能识别算法的自适应集成建模

选择表征建模对象特性漂移的新样本对软测量模型进行自适应更新,能够降低模型复杂度和运行消耗,提高模型可解释性和预测精度.针对新样本近似线性依靠程度（Approximate linear dependence, ALD）和预测误差（Prediction error, PE）等指标只能片面反映建模对象的漂移程度,领域专家结合具体工业过程需要依据上述指标和自身积累经验进行更新样本的有效识别等问题,本文提出了基于更新样本智能识别算法的自适应集成建模策略.首先,基于历史数据离线建立基于改进随机向量泛函连接网络（Improved random vector functional-link networks, IRVFL）的选择性集成模型;然后,基于集成子模型对新样本进行预测输出后采用在线自适应加权算法（On-line adaptive weighting fusion, OLAWF）对集成子模型权重进行更新,实现在线测量阶段对建模对象特性变化的动态自适应;接着基于领域专家知识构建模糊推理模型对新样本相对ALD（Relative ALD, RALD）值和相对PE（Relative PE, RPE）值进行融合,实现更新样本智能识别,构建新的建模样本库;最后实现集成模型的在线自适应更新.采用合成数据仿真验证了所提算法的合理性和有效性.     

基于统计回归的质量推断方法

统计回归是软测量建模的一个重要手段.本文在回顾了典型统计回归方法的基础上,采用部分最小二乘(PLS)建立高定量造纸过程的水分软测量模型.鉴于软测量模型的扰动拟合本质,提出了基于扰动估计的软测量建模策略.水分软测量模型的建立,验证了该建模策略的有效性.     

高速网络流量测量方法

高速网络流量测量是目前实施实时准确地监测、管理和控制网络的基础.基于网络流量测量的应用,将网络流量测量分为抽样方法和数据流方法.从不同的层次,将抽样方法分为分组抽样和流抽样,分别介绍了两类抽样方法;从测度角度介绍了数据流方法.详细介绍了高速网络流量测量的常用数据结构,以及抽样、数据流方法在高速网络流量测量中的应用,比较了各种方法的优劣.概述了高速网络流量测量技术的研究进展.最后,就现有的网络流量测量方法的不足,对网络流量测量的发展趋势和进一步的研究方向进行了讨论.     

减法聚类-ANFIS在网络故障诊断的应用研究

提出了一种基于减法聚类-自适应模糊神经网络（ANFIS）的网络故障诊断建模方法。减法聚类算法生成初始模糊推理系统,ANFIS建立网络故障诊断原始模型,应用混合算法对模糊规则的参数进行训练并建立最终的模型。仿真实验表明基于减法聚类-ANFIS的建模方法是有效的;通过仿真结果比较,减法聚类-ANFIS的网络故障诊断能力及收敛速度均优于BP神经网络,更适合作为网络故障诊断模型。     

测距传感器数据在线自适应加权融合

针对移动机器人的测距系统,采用了红外线传感器与超声波传感器共同测距,避免了因使用单个传感器进行多次测量而降低系统的实时性和产生信号串扰问题;应用自适应加权数据融合估计算法对实时测量数据进行在线融合估计,只对当前采样时刻的测量数据进行自适应加权融合,而各传感器的加权因子则通过传感器的测量数据进行方差在线学习估计以自适应方式进行调整,使融合结果的均方误差始终最小,实现两种传感器在功能上的互补;实验结果表明,该方法提高了整体测距精度,得到了被测距离更加准确的估计.     

Optimal adaptive bandwidth monitoring for QoS based retrieval

Network aware multimedia delivery applications are a class of applications that provide certain level of quality of service (QoS) guarantees to end users while not assuming underlying network resource reservations. These applications guarantee QoS parameters like media object transmission time limit by actively monitoring the available bandwidth of the network and adapting the object to a target size that can be transmitted within a given time limit. A critical problem is how to obtain an accurate enough estimation of available bandwidth while not wasting too much time in bandwidth testing. In this paper, we present an algorithm to determine optimal amount of bandwidth testing given a probabilistic confidence level for network-aware multimedia object retrieval applications. The model treats the bandwidth testing as sampling from an actual bandwidth population. It uses statistical estimation method to quantify the benefit of each new bandwidth-testing sample, which is used to determine the optimal amount of bandwidth testing by balancing the benefit with the cost of each sample. Our implementation and experiments shows the algorithm determines the optimal amount of bandwidth testing effectively with minimum computation overhead.     

使用自负载降速率探针包列的可用带宽测量方法

可用带宽是一条网络链路或路径的剩余带宽,该指标对网络流量工程以及速率敏感应用具有重要意义。现有的可用带宽测量技术大多基于单跳或者流体背景流量模型,在突发背景流及多跳路径下健壮性不足。现提出一种使用自负载降速率探针包列（SLDRT）技术的可用带宽测量方法。基于NS-2的仿真实验表明,SLDRT在多跳路径并基于非流体流量模型的网络环境下比另一重要测量工具pathChirp具有更高的准确性,并且平均测量时间小于30 ms,证明其可以满足高速、高精度的可用带宽测量需求。     

融合多头自注意力机制的无接触心率估计模型

为了在光照变化和头部运动条件下实现准确稳定的无接触心率估计,基于UNet模型提出一种融合多头自注意力机制的端到端心率估计模型rPPG-UNet。该模型通过使用U型编码器—解码器网络结构实现对生理特征的提取与重建,并使用Skip Connection连接编码器与解码器实现浅层时间特征的复用。该模型还融合多头自注意力机制来捕获生理特征的时间依赖性。最后,该模型采用多任务学习策略以提高心率估计的准确度,加速网络训练。在公开数据集上的实验结果表明,rPPG-UNet的性能优于其他基线模型,可以实现更准确的无接触心率估计。     

A VoIP Traffic Identification Scheme Based on Host and Flow Behavior Analysis

With the development of network and multimedia coding techniques, more and more Voice over Internet Protocol (VoIP) applications have emerged. The traffic identification on VoIP applications becomes an important issue in network management and traffic analysis. In this paper, a new traffic identification scheme, which combines traffic flow statistic analysis with host behavior estimation, is proposed to identify the VoIP traffic at the transport layer of the Internet. The host IP addresses and the port numbers are examined as the host behavior to distinguish the VoIP traffic from traditional traffic flows. The packet size has been modeled by a function of entropy while the inter-packet time has been modeled by the self-adaptive estimation. The experiment results show that our scheme could obtain a stable performance. At the same time, the proposed scheme could maintain its validity when existing VoIP applications are updated or the new ones admitted. Both accuracy and flexibility can be improved.     

Online moving horizon estimation of fluxes in metabolic reaction networks

Using online state and parameter estimation, concentrations and fluxes in bioprocesses can be estimated for use in monitoring, optimization and control applications. Existing methodologies, however, either ignore the dynamic nature of the problem, or focus on the extracellular concentration states and pay less attention to accurate flux estimates. These estimates are useful for online monitoring of the flux state of an organism, or for developing novel flux-based strategies for online control of bioreactors.In this contribution, the dynamic metabolic flux analysis model structure is combined with two kinetic flux models: a linear flux model and a nonlinear, more mechanistic flux model. The parameters of these models are estimated online through a moving horizon estimation strategy. The resulting algorithm is illustrated on two simulated case studies: a small-scale network, to assess the influence of important algorithm parameters on the final estimates, and a medium-scale network for Escherichia coli, to empirically test the performance of the methodology in a more realistic situation.An important parameter in this estimation strategy is the chosen noise level on the estimated parameters. This choice is not trivial, but is observed to have a significant influence on the resulting estimates. Furthermore, also the effect of the choice of the null space basis for the stoichiometric matrix of the metabolic reaction network was assessed. In the small-scale case study, it was found that a linear flux model with a specific parameter noise level was performing well for both state and flux estimation. The influence of the choice of the null space basis matrix on the estimation performance was much lower. The resulting scenario was evaluated in the medium-scale case study and found to be performing very well also in that case.     

Determining model accuracy of network traces

Accurate network modeling is critical to the design of network protocols. Traditional modeling approaches, such as Discrete Time Markov Chains (DTMC) are limited in their ability to model time-varying characteristics. This problem is exacerbated in the wireless domain, where fading events create extreme burstiness of delays, losses, and errors on wireless links. In this paper, we describe the data preconditioning modeling technique that is capable of capturing the statistical characteristics of wired and wireless network traces. We revise our previous developed data preconditioning modeling algorithm, the Markov-based Trace Analysis (MTA), and present the Multiple states MTA (MMTA) algorithm. Our main contributions are methodologies created to quantify the accuracy of network models, methodology to choose the most accurate model for a given network and characteristic of interest (e.g., delay, loss, or error process), and the validation of our data preconditioning modeling algorithms.     

基于综合统计法的算法效率分析和优化

针对事前估算法在工程应用中的局限性,提出基于综合统计法的算法效率分析和优化方法。使用多个监测器对算法的运行时间进行事后统计,利用Admahl定律和加速比性能定律指出算法效率的优化方向。与事前估算法相比,该方法无需建立算法和软硬件运行平台的模型,分析简单、结果准确,并能提供具体的优化建议预测优化后算法效率提高的倍数     

A stability assessment of solution adaptation techniques for analogy-based software effort estimation

Among numerous possible choices of effort estimation methods, analogy-based software effort estimation based on Case-based reasoning is one of the most adopted methods in both the industry and research communities. Solution adaptation is the final step of analogy-based estimation, employed to aggregate and adapt to solutions derived during the case-based reasoning process. Variants of solution adaptation techniques have been proposed in previous studies; however, the ranking of these techniques is not conclusive and shows conflicting results, since different studies rank these techniques in different ways. This paper aims to find a stable ranking of solution adaptation techniques for analogy-based estimation. Compared with the existing studies, we evaluate 8 commonly adopted solution techniques with more datasets (12), more feature selection techniques included (4), and more stable error measures (5) to a robust statistical test method based on the Brunner test. This comprehensive experimental procedure allows us to discover a stable ranking of the techniques applied, and to observe similar behaviors from techniques with similar adaptation mechanisms. In general, the linear adaptation techniques based on the functions of size and productivity (e.g., regression towards the mean technique) outperform the other techniques in a more robust experimental setting adopted in this study. Our empirical results show that project features with strong correlation to effort, such as software size or productivity, should be utilized in the solution adaptation step to achieve desirable performance. Designing a solution adaptation strategy in analogy-based software effort estimation requires careful consideration of those influential features to ensure its prediction is of relevant and accurate.     

改进型分层抽样技术及性能研究

报文抽样技术是高速网络流量测量和管理中使用的一项关键技术。本文通过引进分层特征、层数L、分层边界、各层样本量分配、层内抽样策略5个分层抽样参数,并对其进行重新配置和简单理论探讨,实现对分层抽样技术的改进。同时文章使用简单线性估计推断原始流数据,并借助于Φ偏差检验方法,对改进的分层抽样技术和其它抽样技术在测量网络报文长度分布方面进行准确性性能比较。结果表明,改进的分层抽样技术在测量报文长度分布方面的准确性性能远高于其它抽样方式,提高了测量的精度。     

Faster power estimation of CMOS designs using vector compaction - a fractal approach

Low power digital complementary metal oxide semiconductor (CMOS) circuit design requires accurate power estimation. In this paper, we present a compaction algorithm for generating compact vector sets to estimate power efficiently. Power can be estimated using dynamic (simulation) or static (statistical/probabilistic) techniques. Dynamic power estimation techniques simulate the design using a large input vector set for accurate estimation. However, the simulation time is prohibitively long for bigger designs with larger vector sets. The statistical methods, on the other hand, use analytical tools that make them faster but less accurate. To achieve the accuracy of dynamic power estimation and the speed of statistical methods, one approach is to generate a compact, representative vector set that has the same switching transition behavior as the original larger vector set. The compaction algorithm presented in this paper uses fractal concepts to generate such a compact vector set. The fractal technique quantifies correlation by a fractal parameter which can be determined faster than calculating correlation explicitly. Experimental results on circuits from the ISCAS85 and ISCAS89 benchmark suites, with correlated input vector sets, resulted in a maximum compaction ratio of 65.57X (average 38.14X) and maximum power estimation error of 2.4% (average 2.06%). Since the size of the compact vector set used for simulation is smaller, the simulation time will be shorter and will significantly speed up the design cycle.