检索结果-内蒙古大学图书馆

IEEE 3rd International Conference on signal and image processing (ICSIP)

作者： Wu, Xiaoting Feng, Xiaoyi Boutellaa, Elhocine Hadid, Abdenour Univ Oulu Ctr Machine Vis & Signal Anal Oulu Finland Northwestern Polytech Univ Sch Elect & Informat Xian Shaanxi Peoples R China

ISBN: (纸本)9781538663967

Kinship verification from faces is a challenging task that is attracting an increasing attention in the recent years. The proposed methods so far are not robust enough to predict the kin between persons via facial appearance only. The initial studies using deep convolutional neural networks (CNN) have not shown their full potential as well, mainly due to limited training data. To mitigate this problem, we propose a new approach to kinship verification based on color features and extreme learning machines (ELM). While ELM aims to deal with small size training sets, color features are proven to provide significant enhancement over gray-scale counterparts. We evaluate our proposed method on three benchmark and publicly available kinship databases, namely KinFaceW-I, KinFaceW-ii and TSKinFace. The obtained results compares favorably against some state-of-the-art methods including those based on deep learning.

关键词： kinship verification color features extreme learning machine

来源：评论

学校读者我要写书评

暂无评论

CAN MICRO-EXPRESSION BE RECOGNIZED BASED ON SINGLE APEX FRAME? 25

CAN MICRO-EXPRESSION BE RECOGNIZED BASED ON SINGLE APEX FRAM...

引用

25th IEEE International Conference on image processing (ICIP)

作者： Li, Yante Huang, Xiaohua Zhao, Guoying Univ Oulu Ctr Machine Vis & Signal Anal Oulu Finland

ISBN: (纸本)9781479970612

Micro-expressions are rapid and subtle facial movements such that they are difficult to detect and recognize. Most of recent works have attempted to recognize micro-expression by using the spatial and dynamic information from the video clip. Physiological studies have demonstrated that the apex frame can convey the most emotion expressed in facial expression. It may be reasonable to use apex frame for improving micro-expression recognition. However, it is wonder how much apex frames contribute to micro-expression recognition. In this paper, we primarily focus on resolving the contribution-level by using apex frame for micro-expression recognition. Firstly, we propose a new method to detect the apex frame in frequency domain, as it is found that apex frame has very correlated relationship with the amplitude change in frequency domain. Secondly, we propose to use deep convolutional neural network (DCNN) on apex frame to recognize micro-expression. Intensive experimental results on CASME ii database shows that our method has achieved considerably improvement compared with the state-of-the-art methods in micro-expression recognition. These results also demonstrate that apex frame can express the major emotion in micro-expression.

关键词： Micro-expression Apex frame spotting 3D FFT Deep Convolutional neural Network

来源：评论

学校读者我要写书评

暂无评论

A Benchmark Study Of Backdoor Data Poisoning Defenses For Deep neural Network Classifiers And A Novel Defense

A Benchmark Study Of Backdoor Data Poisoning Defenses For De...

引用

IEEE Workshop on Machine Learning for signal processing

作者： Zhen Xiang David J. Miller George Kesidis State College Pennsylvania State University PA USA

While data poisoning attacks on classifiers were originally proposed to degrade a classifier's usability, there has been strong recent interest in backdoor data poisoning attacks, where the classifier learns to classify to a target class whenever a backdoor pattern (e.g., a watermark or innocuous pattern) is added to an example from some class other than the target class. In this paper, we conduct a benchmark experimental study to assess the effectiveness of backdoor attacks against deep neural network (DNN) classifiers for images (CIFAR-10 domain), as well as of anomaly detection defenses against these attacks, assuming the defender has access to the (poisoned) training set. We also propose a novel defense scheme (cluster impurity (CI)) based on two ideas: i) backdoor patterns may cluster in a DNN's (e.g. penultimate) deep layer latent space; ii) image filtering (or additive noise) may remove the backdoor patterns, and thus alter the class decision produced by the DNN. We demonstrate that largely imperceptible single-pixel backdoor attacks are highly successful, with no effect on classifier usability. However, the CI approach is highly effective at detecting these attacks, and more successful than previous backdoor detection methods.

关键词： Training Feature extraction Impurities Neurons neural networks Benchmark testing Usability

来源：评论

学校读者我要写书评

暂无评论

Automatic Facial Age Estimate Based on Convolution neural Network 13th

Automatic Facial Age Estimate Based on Convolution Neural Ne...

引用

13th International Conference on Intelligent Information Hiding and Multimedia signal processing (iiH-MSP)

作者： Zou, Jiancheng Yang, Xuan Zhang, Honggen Chen, Xiaoguang North China Univ Technol Inst Image Proc & Pattern Recognit Beijing 100144 Peoples R China

ISBN: (纸本)9783319638591;9783319638584

As a new biometric recognition technology, automatic age estimation based on facial image has become an important subject in the field of computer vision and human-computer interaction (HCI). And automatic facial age estimate system has been increasingly used in criminal investigation, image retrieval and intelligent monitoring in recent years. Therefore, the research of facial age estimation has a broad prospect. Convolution neural network (CNN) as a deep learning architecture can extract the essential features of the facial image with a better effect than traditional methods, especially in the case of large changes in imaging shooting conditions. In this paper, an improved method of facial age estimate based on CNN is proposed. By considering the number limitation of the existing age estimation data sets, we adopt the method of fine-tuning the existed network model. The recognition rate can be increased by 3% based on the proposed method. A facial age estimate system has been constructed for applications and the experimental results show that the system can meet the real-time application needs.

关键词： Facial age estimation CNN Deep learning

来源：评论

学校读者我要写书评

暂无评论

A Deep Learning Based Alternative to Beamforming Ultrasound images

A Deep Learning Based Alternative to Beamforming Ultrasound ...

引用

2018 IEEE International Conference on Acoustics, Speech, and signal processing, ICASSP 2018

作者： Nair, Arun Asokan Tran, Trac D. Reiter, Austin Lediju Bell, Muyinatu A. Department of Electrical and Computer Engineering Johns Hopkins University Baltimore United States Department of Computer Science Johns Hopkins University Baltimore United States Department of Biomedical Engineering Johns Hopkins University Baltimore United States

ISBN: (纸本)9781538646588

Deep learning methods are capable of performing sophisticated tasks when applied to a myriad of artificial intelligent (AI) research fields. In this paper, we introduce a novel approach to replace the inherently flawed beamforming step during ultrasound image formation by applying deep learning directly to RF channel data. Specifically, we pose the ultrasound beamforming process as a segmentation problem and apply a fully convolutional neural network architecture to segment anechoic cysts from surrounding tissue. We train our network on a dataset created using the Field ii ultrasound simulation software to simulate plane wave imaging with a single insonification angle. We demonstrate the success of our architecture in extracting tissue information directly from the raw channel data, which completely bypasses the beamforming step that would otherwise require multiple insonifi-cation angles for plane wave imaging. Our simulated results produce mean Dice coefficient of 0.98 ± 0.02, when measuring the overlap between ground truth cyst locations and cyst locations determined by the network. The proposed approach is promising for developing dedicated deep-learning networks to improve the real-time ultrasound image formation process. © 2018 IEEE.

关键词： Ultrasonic imaging Imaging Array signal processing neural networks Radio frequency image segmentation Machine learning"

来源：评论

学校读者我要写书评

暂无评论

Space-time error estimates for deep neural network approximations for differential equations

arXiv

引用

arXiv 2019年

作者： Grohs, Philipp Hornung, Fabian Jentzen, Arnulf Zimmermann, Philipp Faculty of Mathematics and Research Platform Data Science University of Vienna Austria Department of Mathematics ETH Zurich Zürich Switzerland Faculty of Mathematics Karlsruhe Institute of Technology Karlsruhe Germany Department of Mathematics ETH Zurich Zürich Switzerland Department of Mathematics ETH Zurich Zürich Switzerland Faculty of Mathematics University of Regensburg Regensburg Germany

Over the last few years deep artificial neural networks (DNNs) have very successfully been used in numerical simulations for a wide variety of computational problems including computer vision, image classification, speech recognition, natural language processing, as well as computational advertisement. In addition, it has recently been proposed to approximate solutions of partial differential equations (PDEs) by means of stochastic learning problems involving DNNs. There are now also a few rigorous mathematical results in the scientific literature which provide error estimates for such deep learning based approximation methods for PDEs. All of these articles provide spatial error estimates for neural network approximations for PDEs but do not provide error estimates for the entire space-time error for the considered neural network approximations. It is the subject of the main result of this article to provide space-time error estimates for DNN approximations of Euler approximations of certain perturbed differential equations. Our proof of this result is based (i) on a certain artificial neural network (ANN) calculus and (ii) on ANN approximation results for products of the form r0, T s × Rd Q pt, xq ÞÑ tx P Rd where T P p0, 8q, d P N, which we both develop within this *** Codes 65L70, 68T99 Copyright © 2019, The Authors. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Action Recognition Using the 3D Dense Micro-Block Difference 2

Action Recognition Using the 3D Dense Micro-Block Difference

引用

Conference on Counterterrorism, Crime Fighting, Forensics, and Surveillance Technologies ii

作者： Voronin, V. Pismenskova, M. Zelensky, A. Cen, Yigang Nadykto, A. Egiazarian, K. Don State Tech Univ Lab Math Methods Image Proc & Intelligent Comp Vi Rostov Na Donu Russia Moscow State Univ Technol STANKIN Moscow Russia Beijing Jiaotong Univ Inst Informat Sci Beijing Peoples R China Tampere Univ Technol Dept Signal Proc Tampere Finland

ISBN: (数字)9781510621886

ISBN: (纸本)9781510621886

This paper describes a framework for action recognition which aims to recognize the goals and activities of one or more human from a series of observations. We propose an approach for the human action recognition based on the 3D dense micro-block difference. The proposed algorithm is a two-stage procedure: (a) image preprocessing using a 3D Gabor filter and (b) a descriptor calculation using 3D dense micro-block difference with SVM classifier. At the first step, an efficient spatial computational scheme designed for the convolution with a bank of 3D Gabor filters is present. This filter intensifies motion using a convolution for a set of 3D patches and arbitrarily-oriented anisotropic Gaussian. For preprocessed frames, we calculate the local features such as 3D dense micro-block difference (3D DMD), which capture the local structure from the image patches at high scales. This approach is processing the small 3D blocks with different scales from frames which capture the microstructure from it. The proposed image representation is combined with fisher vector method and linear SVM classifier. We evaluate the proposed approach on the UCF50, HMDB51 and UCF101 databases. Experimental results demonstrate the effectiveness of the proposed approach on video with a stochastic textures background with comparisons of the state-of-the-art methods.

关键词： action recognition texture micro-block difference 3D Gabor filter

来源：评论

学校读者我要写书评

暂无评论

EFFICIENT CYCLIC LEARNING RATE SCHEDULES AND THEIR EVALUATIONS FOR neural NETWORK ENSEMBLE

EFFICIENT CYCLIC LEARNING RATE SCHEDULES AND THEIR EVALUATIO...

引用

IEEE Workshop on Machine Learning for signal processing

作者： Minoru Mori Masahiro Nakano Faculty of Information Technology Kanagawa Institute of Technology Japan NTT Communication Science Laboratories NTT Corporation Japan

Warm restart techniques on training deep neural networks often achieve better recognition accuracies and can be regarded as easy methods to obtain multiple neural networks with no additional training cost from a single training process. Ensembles of intermediate neural networks obtained by warm restart techniques can provide higher accuracy than a single neural network obtained finally by a whole training process. However, existing methods on both of warm restart and its ensemble techniques use fixed cyclic schedules and have little degree of parameter adaption. This paper extends a class of possible schedule strategies of warm restart, and clarifies their effectiveness for recognition performance. Specifically, we propose parameterized functions and various cycle schedules to improve recognition accuracies by the use of deep neural networks with no additional training cost. Experiments on CIFAR-10 and CIFAR-100 show that our methods can achieve more accurate rates than the existing cyclic training and ensemble methods.

关键词： neural networks Training Schedules image recognition Speech recognition stochastic processes Optimization

来源：评论

学校读者我要写书评

暂无评论

Real-Time Denoising of Brillouin Optical Time Domain Analyzer With High Data Fidelity Using Convolutional neural Networks

引用

Journal of Lightwave Technology 2018年第11期37卷 2648-2653页

作者： Hao Wu Yangyang Wan Ming Tang Yunjin Chen Can Zhao Ruolin Liao Yiqing Chang Songnian Fu Perry Ping Shum Deming Liu Yangyang Wan Ming Tang Yunjin Chen Can Zhao Ruolin Liao Yiqing Chang Songnian Fu Perry Ping Shum Deming Liu 2648

In recent years, many conventional image denoising techniques have been intensively studied to enhance the signal to noise ratio (SNR) of Brillouin optical time domain analyzer (BOTDA), due to their superior denoising performance to one-dimensional methods. However, in the case of low sampling rate, the details of the signal are smoothed out due to less useful information, resulting in a degradation of the spatial resolution. Moreover, these conventional denoising algorithms are quite time-consuming compared with the BOTDA measuring time. To overcome these drawbacks, we employ a feed-forward convolutional neural networks (CNN) based image denoising for BOTDA. A conventional BOTDA system with 15 ns pulse width is implemented to demonstrate the effectiveness of the exploited CNN-based denoising method. The actual electrical noise signals of the BOTDA at different sampling rates are collected to synthesize training samples. The CNN model is trained with the noise and simulated BOTDA signals. Experimental results show that SNR improvement of 13.43 dB, 13.57 dB, and 12.9 dB is achieved at a sampling rate of 500 MSa/s, 250 MSa/s, and 125 MSa/s, respectively, via the trained CNN denoiser. No spatial resolution distortion can be observed in the denoised BOTDA signals. Besides, the CNN denoiser only takes 0.045 s to process a 151 × 50000 image benefiting from GPU computing. This processing time is negligible compared with the acquisition time of BOTDA, which makes real-time denoising possible.

关键词： Fiber optic cables neural networks Optical signals signal processing Spatial resolution stochastic gradient descent

来源：评论

学校读者我要写书评

暂无评论

TDOA-based Localization via stochastic Gradient Descent Variants

TDOA-based Localization via Stochastic Gradient Descent Vari...

引用

IEEE Vehicular Technology Conference

作者： Luis F. Abanto-Leon Arie Koppelaar Sonia Heemstra de Groot Eindhoven University of Technology NXP Semiconductors

ISBN: (纸本)9781538663592;9781538663585

Source localization is of pivotal importance in several areas such as wireless sensor networks and Internet of Things (IoT), where the location information can be used for a variety of purposes, e.g. surveillance, monitoring, tracking, etc. Time Difference of Arrival (TDOA) is one of the well-known localization approaches where the source broadcasts a signal and a number of receivers record the arriving time of the transmitted signal. By means of computing the time difference from various receivers, the source location can be estimated. On the other hand, in the recent few years novel optimization algorithms have appeared in the literature for (i) processing big data and for (ii) training deep neural networks. Most of these techniques are enhanced variants of the classical stochastic gradient descent (SGD) but with additional features that promote faster convergence. In this paper, we compare the performance of the classical SGD with the novel techniques mentioned above. In addition, we propose an optimization procedure called RM-SProp+AF, which is based on RMSProp algorithm but with the advantage of incorporating adaptation of the decaying factor. We show through simulations that all of these techniques--which are commonly used in the machine learning domain--can also be successfully applied to signal processing problems and are capable of attaining improved convergence and stability. Finally, it is also shown through simulations that the proposed method can outperform other competing approaches as both its convergence and stability are superior.

关键词： gradient methods greedy algorithms Internet of Things learning (artificial intelligence) neural nets optimisation signal processing stochastic processes time-of-arrival estimation wireless sensor networks time-of-arrival estimation Internet of things neural network greedy algorithm gradient methods Wireless sensor networks Markov chain signal processing Position measurement POSITION INFORMATION Optimization Procedure stochastic Processes Time difference of arrival

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：