检索结果-内蒙古大学图书馆

第十四届全国人机语音通讯学术会议

作者： Jiangyan Yi Jianhua Tao Zhengqi Wen Ya Li Hao Ni National Laboratory of Pattern Recognition Institute of AutomationChinese Academy of Sciences School of Computer and Control Engineering University of Chinese Academy of Sciences CAS Center for Excellence in Brain Science and Intelligence Technology Institute of AutomationChinese Academy of Sciences

Mobile devices have limited computing power and limited ***,large deep neural network(DNN) based acoustic models are not well suited for application on mobile *** order to alleviate this problem,this paper proposes to compress acoustic models by using knowledge *** approach forces a large teacher model to transfer generalized knowledge to a small student *** student model is trained with a linear interpolation of hard probabilities and soft probabilities to learn generalized knowledge from the teacher *** hard probabilities are generated from a Gaussian mixture model hidden Markov model(GMM-HMM) *** soft probabilities are computed from a teacher model(DNN or RNN).Experiments on AMI corpus show that a small student model obtains 2.4%relative WER improvement over a large teacher model with almost 7.6 times compression ratio.

关键词： model compression knowledge transfer deep neural networks automatic speech recognition

来源：评论

学校读者我要写书评

暂无评论

On active decoupling control for horizontal-lateral direction of hypersonic flight based on HIFOO 21

On active decoupling control for horizontal-lateral directio...

引用

21st AIAA International Space Planes and Hypersonics Technologies Conference, Hypersonics 2017

作者： Ma, Shunjian Zhu, Kai Sun, Mingwei Wang, Zenghui Chen, Zengqiang Dept. of Automation and Intelligent Science Nankai University College of Computer and Control Engineering Tianjin300350 China Beijing Aerospace Technology Institute Tianjin Key Laboratory of Intelligent Robotics Tianjin300350 China Beijing Aerospace Technology Institute Beijing100074 China Department of Electrical and Mining Engineering University of South Africa FL1710 South Africa

ISBN: (数字)9781624104633

ISBN: (纸本)9781624104633

Nowadays, hypersonic flight vehicles are widely studied. Since their high nonlinearity, strong coupling and uncertain factors, the conventional three-channel independent control design method is often unable to work well. In this paper, an active disturbance rejection controller (ADRC) is utilized to cancel internal uncertainties and exogenous disturbances for longitudinal dynamics. Then, the MIMO robust controller is obtained by choosing proper sensitivity functions and weighting functions to reduce horizontal-lateral direction coupling. The conventional H robust controllor is high-order, which is difficult for practitioners to implement. With frequency transformation, both ADRC for the longitudinal dynamics and MIMO robust controller for the horizontal-lateral dynamics are reformed into the static output feedback structure to reduce the high order of controller. Finally, all parameters in the controllers can be optimized by HIFOO (H-Infinity Fixed Order Optimization). The Winged-Cone, a 6-freedom NASA hypersonic vehicle model, is used as a benchmark, and Monte Carlo simulations demonstrate the effectiveness of the proposed method. © 2017, American institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.

关键词： controllers

来源：评论

学校读者我要写书评

暂无评论

On the nonexistence of event-based triggers that preserve Gaussian state in presence of package-drop

On the nonexistence of event-based triggers that preserve Ga...

引用

American control Conference

作者： Enoch Kung Junfeng Wu Dawei Shi Ling Shi Electronic and Computer Engineering Hong Kong University of Science and Technology Department of Automatic Control The Royal Institute of Technology (KTH) State Key Laboratory of Intelligent Control and Decision of Complex Systems School of Automation Beijing Institute of Technology

ISBN: (纸本)9781509045839

State estimation is a core objective in cyber-physical systems. In the state estimation problem over linear systems, the Kalman filter is the standard solution. The filter is the format on which the solutions to subsequent estimation problems are based. Among these problems are the estimation problem in the presence of packet drops and estimation problem involving event-based triggers. We study in this paper both phenomena simultaneously. In an attempt to find the Kalman-like filter, which proves the Gaussianity of the state and offers a set of update equations, our paper shows that no such filter exists. More precisely, one cannot find an event-based trigger such that under possible packet drops, the state variable remains a Gaussian variable. This conclusion can be reasonably extended to a more general setting.

关键词： State estimation Packet drops Event-based triggers State estimation Event-based Filters State variable TRIGGERS Gauss

来源：评论

学校读者我要写书评

暂无评论

Dielectric cross-shaped resonator based metasurface for vortex beam generation in Mid-IR and THz wavelengths

arXiv

引用

arXiv 2019年

作者： Dharmavarapu, Raghu Izumi, Ken-Ichi Katayama, Ikufumi Ng, Soon Hock Vongsvivut, Jitraporn Tobin, Mark J. Kuchmizhak, Aleksandr Nishijima, Yoshiaki Bhattacharya, Shanti Juodkazis, Saulius Department of Electrical Engineering Indian Institute of Technology Madras Chennai600036 India Centre for Micro-Photonics Faculty of Science Engineering and Technology Swinburne University of Technology HawthornVIC3122 Australia Melbourne Centre for Nanofabrication ANFF 151 Wellington Road ClaytonVIC3168 Australia Physics Department Yokohama National University 79-5 Tokiwadai Hodogaya-ku Yokohama240-8501 Japan Infrared Microspectroscopy Beamline Australian Synchrotron ClaytonVIC3168 Australia Institute of Automation and Control Processes Far Eastern Branch Russian Academy of Sciences Vladivostok690041 Russia Far Eastern Federal University Vladivostok690090 Russia Department of Electrical and Computer Engineering Graduate School of Engineering Yokohama National University 79-5 Tokiwadai Hodogaya-ku Yokohama240-8501 Japan Institute of Advanced Sciences Yokohama National University 79-5 Tokiwadai Hodogaya-ku Yokohama240-8501 Japan School of Materials and Chemical Technology Tokyo Institute of Technology 2-12-1 Ookayama Meguro-ku Tokyo152-8550 Japan

Metasurfaces are engineered thin surfaces comprising two dimensional (2D) arrays of sub-wavelength spaced and sub-wavelength sized resonators. Metasurfaces can locally manipulate the amplitude, phase and polarization of light with high spatial resolution. In this study, we report numerical and experimental results of a vortex-beam-generating metasurface fabricated specifically for infrared (IR) and terahertz (THz) wavelengths. The designed metasurface consisted of a 2D array of dielectric cross-shaped resonators with spatially varying length, thereby providing desired spatially varying phase shift to the incident light. The metasurface was found to be insensitive to polarization of incident light. The dimensions of the cross-resonators were calculated using rigorous finite difference time domain (FDTD) analysis. The spectral scalability via physical scaling of meta resonators was demonstrated using two vortex generating optical elements operating at 8.8 μm (IR) and 0.78 THz (Terahertz). The vortex beam generated in the mid-IR spectral range was imaged using FTIR imaging miscroscope equipped with a focal plane array (FPA) detector. This design could be used for efficient wavefront shaping as well as various optical imaging applications in mid-IR spectral range, where polarization insensitivity is desired. Copyright © 2019, The Authors. All rights reserved.

关键词： Resonators

来源：评论

学校读者我要写书评

暂无评论

Body joint guided 3D deep convolutional descriptors for action recognition

arXiv

引用

arXiv 2017年

作者： Cao, Congqi Zhang, Yifan Zhang, Chunjie Lu, Hanqing National Laboratory of Pattern Recognition Institute of Automation Chinese Academy of Sciences Beijing100190 China School of Computer and Control Engineering University of Chinese Academy of Sciences

Three dimensional convolutional neural networks (3D CNNs) have been established as a powerful tool to simultaneously learn features from both spatial and temporal dimensions, which is suitable to be applied to video-based action recognition. In this work, we propose not to directly use the activations of fully-connected layers of a 3D CNN as the video feature, but to use selective convolutional layer activations to form a discriminative descriptor for video. It pools the feature on the convolutional layers under the guidance of body joint positions. Two schemes of mapping body joints into convolutional feature maps for pooling are discussed. The body joint positions can be obtained from any off-the-shelf skeleton estimation algorithm. The helpfulness of the body joint guided feature pooling with inaccurate skeleton estimation is systematically evaluated. To make it end-to-end and do not rely on any sophisticated body joint detection algorithm, we further propose a two-stream bilinear model which can learn the guidance from the body joints and capture the spatio-temporal features simultaneously. In this model, the body joint guided feature pooling is conveniently formulated as a bilinear product operation. Experimental results on three real-world datasets demonstrate the effectiveness of body joint guided pooling which achieves promising performance. Copyright © 2017, The Authors. All rights reserved.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Adaptive droop control with self-adjusted virtual impedance for three-phase inverter under unbalanced conditions 50

Adaptive droop control with self-adjusted virtual impedance ...

引用

50th IEEE International Symposium on Circuits and Systems, ISCAS 2017

作者： Lu, Zelun Li, Wenxuan Li, Zhen Chen, Xi Lu, Herbert H. C. Dong, Ning Liu, Xiangdong School of Automation Key Laboratory for Intelligent Control and Decision on Complex Systems Beijing Institute of Technology BeijingI0008I China Global Energy Interconnection Research Institute North America Santa Clara 5451 Great America Parkway CA95054 United States School of Electrical Electrical and Computer Engineering University of Western Australia CrawleyWA6009 Australia

ISBN: (纸本)9781467368520

Three-phase inverter is a very important interface in microgrid. Droop control and virtual impedance are widely used to improve the power sharing capability and stability of such system, which however becomes ineffective under unbalanced conditions and may even cause potential stability problems. This paper proposes a fully self-adjusted virtual impedance design to guarantee the decoupling of active and reactive power. The arbitrary unbalanced information is acquired through the pseudo-inverse impedance matrix modeled by full-dq technique. Based on the unbalanced impedance identified, the adaptive droop control is achieved for not only accurate power sharing but also reliable voltage support with autonomous unbalanced voltage compensation. Simulations results verify the accurate identification of unbalanced impedance and effectiveness of the proposed method on the unbalanced compensation. © 2017 IEEE.

关键词： Electric inverters

来源：评论

学校读者我要写书评

暂无评论

Breast Cancer Diagnosis Using an Unsupervised Feature Extraction Algorithm Based on Deep Learning

Breast Cancer Diagnosis Using an Unsupervised Feature Extrac...

引用

第37届中国控制会议

作者： Yawen Xiao Jun Wu Zongli Lin Xiaodong Zhao Department of Automation Shanghai Jiao Tong Universityand Key Laboratory of System Control and Information Processing of Ministry of Education The Center for Bioinformatics and Computational Biology Shanghai Key Laboratory of Regulatory Biologythe Institute of Biomedical Sciences and School of Life SciencesEast China Normal University Charles L.Brown Department of Electrical and Computer Engineering University of Virginia School of Biomedical Engineering Shanghai Jiao Tong University

Breast cancer is the most often detected cancer in women. At the same time, it is one of the most curable types of cancer if diagnosed early. With the development of the detection technology, a growing amount of clinical data and highdimensional features can be used for breast cancer diagnosis. The high-dimensional data contributes to advances in the diagnostic technology, but also incurs a large amount of computational redundancy. Thus, extracting important information and reducing the feature dimension is critical to effective prediction and an accurate treatment ***, the previous works for breast cancer diagnosis are mainly based on labeled data that is difficult to obtain. To address this issue, in this paper, we demonstrate a new scheme, which integrates a deep learning based unsupervised feature extraction algorithm, the stacked auto-encoders, with a support vector machine model（SAE-SVM）, for breast cancer diagnosis. The stacked auto-encoders with the greedy layerwise pre-training and an improved momentum update algorithm is applied to capture essential information and extract necessary features of the original data. Then, a support vector machine model is employed to classify the samples with new features into malignant or benign tumors. The proposed method was tested on the Wisconsin Diagnostic Breast Cancer data set. The performance is evaluated using various measures and compared with the previously published results. The comparison results show that the proposed SAE-SVM method improves the accuracy to 98.25% and outperforms the other methods. The deep learning based unsupervised feature extraction significantly improves the performance of classification and provides a promising approach to breast cancer diagnosis.

关键词： Deep learning auto-encoder feature extraction breast cancer diagnosis unlabeled data

来源：评论

学校读者我要写书评

暂无评论

Motion Detection Based on Improved Sobel and ViBe Algorithm 35

Motion Detection Based on Improved Sobel and ViBe Algorithm

引用

第35届中国控制会议

作者： ZHANG Yujia ZHAO Xiaoguang TAN Min State Key Laboratory of Management and Control for Complex Systems Institute of AutomationChinese Academy of Sciences University of Chinese Academy of Sciences School of Computer and Control Engineering

ISBN: (纸本)9781509009107

Motion detection plays a crucial role in most video based applications.A particular background subtraction technique called ViBe(Visual Background Extractor) is commonly used to obtain foreground objects from the background due to its high detection rate and low computational ***,the performance is not very ***,this paper presents an improved ViBe algorithm to increase the accuracy and robustness of motion ***,a foreground feature map is created by optimizing the result of ViBe *** the edge detection of the original video frames is achieved after pre-sharpening using improved Sobel operator and Otsu ***,by feature fusion(of the foreground and background feature maps) and contour filling,the motion detection results can be *** experiments demonstrate the improvements of the proposed modifications at a limited additional cost.

关键词： Motion detection Otsu algorithm Sobel operator Visual Background Extractor(ViBe)

来源：评论

学校读者我要写书评

暂无评论

Nonlinear observability of grinding mill conditions

引用

IFAC-PapersOnLine 2016年第20期49卷 13-18页

作者： le Roux, J.D. Steinboeck, A. Kugi, A. Craig, I.K. Department of Electrical Electronic and Computer Engineering University of Pretoria Pretoria South Africa Automation and Control Institute Vienna University of Technology Gußhausstraße 27-29 Vienna1040 Austria Christian Doppler Laboratory for Model-Based Control in the Steel Industry Automation and Control Institute Vienna University of Technology Gußhausstraße 27-29 Vienna1040 Austria

A nonlinear observer model of a grinding mill is developed. The model distinguishes between the volumetric hold-up of water, solids, rocks and balls in the mill, where solids are all ore small enough to discharge through the end-discharge grate, and rocks are all ore too large to discharge. The model includes the discharge rate, abrasion rate of rocks and the abrasion rate of balls as parameters. It is shown that with mill discharge flow-rate, discharge density and volumetric hold-up measurements the model states and parameters are locally (weakly) nonlinearly observable, but not linearly observable. Although instrumentation at the mill discharge is not yet included in industrial circuits because of space restrictions, this study motivates the benefits to be gained from including instrumentation at the mill discharge. If mills are fitted with adequate measurement instrumentation at the discharge, important mill states and parameters can be estimated. © 2016

关键词： Ball mills Abrasion Comminution Flow rate Grinding (comminution) Grinding (machining) Grinding mills Observability Parameter estimation State estimation Abrasion rates Discharge density Discharge flow rates Discharge rates Industrial circuits Measurement instrumentation Non linear observer Nonlinear observability

来源：评论

学校读者我要写书评

暂无评论

Bird's Eye View on Systems and control - General View and Case Studies

引用

IFAC-PapersOnLine 2016年第6期49卷 274-279页

作者： Vámos, T. Bars, R. Sik, D. Institute of Computer Science and Control Hungarian Academy of Sciences MTA-SZTAKI Budapest Hungary Dept. of Automation and Applied Informatics Budapest University of Technology and Economics MTA-BME Control Engineering Research Group Budapest Hungary

Systems and control are all around us. System view, understanding systems and how they are controlled is important for everyone. The behavior of systems is determined by some fundamental principles which can be understood by everyone. The main principles governing systems and control are presented on different levels. The first level is dedicated for everyone. Comics-like interpretation is provided with explanations. The second level discusses the main topics dedicated for the interested students. Several case studies demonstrate how system view is considered in different areas of life. With this approach systems in different areas of life could be better understood contributing to better decisions influencing their performance. The third level provides deeper insight into some aspects of modern mathematical system theory, to geometric view of control systems. The material is available on a website on the internet. © 2016

关键词： control theory Algorithms Case studies control disciplines control education control structure Systems and control

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：