一种基于卷积神经网络的性别识别方法

采用人工智能进行性别识别时,人脸图像在获取的时候容易受到光照、遮挡等影响,这些因素给人脸性别识别带来了困难。采用卷积神经网络用于性别识别,并通过扩展网络结构,进一步增强卷积神经网络的分类能力。并且对识别效果进行置信度分析,通过设置卷积神经网络的拒识区域来解决拒绝区间的问题。在实际测试中,通过拒绝7.46%的测试样本,达到98.67%的正确识别率。     

A dynamic cross-layer data queue management approach based on priority for delay-tolerant mobile sensor networks

This paper presents a Dynamic Cross-layer Data Queue Management approach (DC-DQM) based on priority to address the priority deviation problem in Delay-Tolerant Mobile Sensor Networks (DT-MSNs). Receiver-driven data delivery scheme is used for fast response to data transfers, and a priority based interaction model is adopted to identify the data priority. Three interactive parameters are introduced to prioritize and dynamically manage data queue. The experimental results show that it can ameliorate data delivery ratio and achieve good performance in terms of average delay.     

Dynamic proxy tree-based data dissemination schemes for wireless sensor networks

In wireless sensor networks, efficiently disseminating data from a dynamic source to multiple mobile sinks is important for the applications such as mobile target detection and tracking. The tree-based multicasting scheme can be used. However, because of the short communication range of each sensor node and the frequent movement of sources and sinks, a sink may fail to receive data due to broken paths, and the tree should be frequently reconfigured to reconnect sources and sinks. To address the problem, we propose a dynamic proxy tree-based framework in this paper. A big challenge in implementing the framework is how to efficiently reconfigure the proxy tree as sources and sinks change. We model the problem as on-line constructing a minimum Steiner tree in an Euclidean plane, and propose centralized schemes to solve it. Considering the strict energy constraints in wireless sensor networks, we further propose two distributed on-line schemes, the shortest path-based (SP) scheme and the spanning range-based (SR) scheme. Extensive simulations are conducted to evaluate the schemes. The results show that the distributed schemes have similar performance as the centralized ones, and among the distributed schemes, the SR scheme outperforms the SP scheme.     

采用量子神经网络的音频水印新算法

提出了一种采用小波变换和量子神经网络的音频数字水印算法。首先对分帧的音频信号进行小波分解,利用量子神经网络将音频信号的小波低频系数映射为数字水印;然后利用分类准确小波低频系数替换少量分类模糊的小波低频系数,提高水印检测正确率。实验结果表明,通过合理选择替换门限,可以提高算法的鲁棒性,有效抵御噪声、低通滤波、重采样、重量化等攻击。在无门限条件下,相比BP神经网络的水印检测正确率平均提高约1%。      

ON THE STABILITY OF CELLULAR NEURAL NETWORKS WITH FEEDBACK MODE

Cellular Neural Networks (CNN) with feedback mode and M×N cells are equivalent to a network which possesses 2M×N cells, a neighborhood with mirror-like structure, space-variant templates and without feedback as well as without input templates. The stability of the CNN with feedback mode and transformations with the neighborhood of mirror-like structure are discussed.     

A NEW METHOD FOR EIGENSTRUCTURES EXTRACTION AND ITS NEURAL NETWORKS IMPLEMENTATION

The cost function for eigenstructures extraction is discussed in detail in this paper, one can obtain the largest eigenvector by minimizing the cost function. In order to obtain other eigenvectors, a covariance matrix series is constructed. If one compares the cost function with the energy function of a neural networks, the neural networks can be easily introduced to extract the eigenvectors. Theoretical analysis and computer simulations show that the proposed method is reasonable and feasible.     

基于神经网络的混沌信号源的设计及同步

该文应用具有全局最优的BP改进算法和神经网络的强大学习能力、逼近任意非线性能力和权值调整的灵活性来优化混沌信号源的设计,采用非线性负反馈实现了神经网络混沌信号源之间的同步。计算机仿真结果表明:由于该模型充分利用了逼近任意非线性能力和网络权值调整的灵活性,比单一混沌映射能产生更多的、具有良好相关性能的混沌信号,且易于同步。     

Defense against collusion scheme based on elliptic curve cryptography for wireless sensor networks

Wireless Sensor Networks (WSNs) are being deployed for a wide variety of applications and the security problems of them have received considerable attention. Considering the limitations of power, com-putation capability and storage resources, this paper proposed an efficient defense against collusion scheme based on elliptic curve cryptography for wireless sensor networks in order to solve the problems that sensor node-key leaking and adversaries make compromised nodes as their collusions to launch new attack. In the proposed scheme, the group-key distribution strategy is employed to compute the private key of each sensor node, and the encryption and decryption algorithms are constructed based on Elliptic Curve Cryptography (ECC). The command center (node) only needs to broadcast a controlling header with three group elements, and the authorized sensor node can correctly recover the session key and use it to decrypt the broadcasting message. Analysis and proof of the proposed scheme’s efficiency and security show that the proposed scheme can resist the k-collusion attack efficiently.     

Secure data aggregation without persistent cryptographic operations in wireless sensor networks

In-network data aggregation is an essential operation to reduce energy consumption in large-scale wireless sensor networks. With data aggregation, however, raw data items are invisible to the base station and thus the authenticity of the aggregated data is hard to guarantee. A compromised sensor node may forge an aggregation value and mislead the base station into trusting a false reading. Due to the stringent constraints of energy supply and computing capability on sensor nodes, it is challenging to detect a compromised sensor node and keep it from cheating, since expensive cryptographic operations are unsuitable for tiny sensor devices. This paper proposes a secure aggregation tree (SAT) to detect and prevent cheating. Our method is essentially different from other existing solutions in that it does not require any cryptographic operations when all sensor nodes work honestly. The detection of cheating is based on the topological constraints in the aggregation tree. We also propose a weighted voting scheme to determine a misbehaving node and a secure local recovery scheme to avoid using the misbehaving node.     

基于多目标优化的无线传感器网络移动充电及数据收集算法

近年来,通过引入移动设备(ME)为无线传感器网络(WSNs)进行无线充电和数据收集成为一个研究热点。传统方法一般先根据节点的充电需求优先级确定移动路径,再根据该路径依次对节点进行数据收集。该文同时考虑充电需求和数据收集两个维度,以最大化ME的总能量利用率和最小化数据收集平均时延为目标,建立多目标一对多充电及数据收集模型。在ME携带的行驶能量和充电能量不足的前提下,设计路径规划策略和均衡化充电策略,并改进多目标蚁群算法对该文问题进行求解。实验结果表明,该文算法在多种场景下的目标值、Pareto解的数量、Pareto解集的均匀性、分布范围等性能指标均优于NSGA-II算法。     

Energy balanced data propagation in wireless sensor networks

We study the problem of energy-balanced data propagation in wireless sensor networks. The energy balance property guarantees that the average per sensor energy dissipation is the same for all sensors in the network, during the entire execution of the data propagation protocol. This property is important since it prolongs the network’:s lifetime by avoiding early energy depletion of sensors. We propose a new algorithm that in each step decides whether to propagate data one-hop towards the final destination (the sink), or to send data directly to the sink. This randomized choice balances the (cheap) one-hop transimssions with the direct transimissions to the sink, which are more expensive but “bypass” the sensors lying close to the sink. Note that, in most protocols, these close to the sink sensors tend to be overused and die out early. By a detailed analysis we precisely estimate the probabilities for each propagation choice in order to guarantee energy balance. The needed estimation can easily be performed by current sensors using simple to obtain information. Under some assumptions, we also derive a closed form for these probabilities. The fact (shown by our analysis) that direct (expensive) transmissions to the sink are needed only rarely, shows that our protocol, besides energy-balanced, is also energy efficient. This work has been partially supported by the IST/FET/GC Programme of the European Union under contract numbers IST-2001-33135 (CRESCCO) and 6FP 001907 (DELIS). A perliminary version of the work appeared in WMAN 2004 [11]. Charilaos Efthymiou graduated form the Computer Engineering and Informatics Department (CEID) of the University of Patras, Greece. He received his MSc from the same department with advisor in S. Nikoletseas. He currently continuous his Ph.D studies in CEID with advisor L. Kirousis. His research interest include Probabilistic Techniques and Random Graphs, Randomized Algorithms in Computationally Hard Problems, Stochastic Processes and its Applications to Computer Science. Dr. Sotiris Nikoletseas is currently a Senior Researcher and Managing Director of Research Unit 1 (“Foundations of Computer Science, Relevant Technologies and Applications”) at the Computer Technology Institute (CTI), Patras, Greece and also a Lecturer at the Computer Engineering and Informatics Department of Patras University, Greece. His research interests include Probabilistic Techniques and Random Graphs, Average Case Analysis of Graph Algorithms and Randomized Algorithms, Fundamental Issues in Parallel and Distributed Computing, Approximate Solutions to Computationally Hard Problems. He has published scientific articles in major international conferences and journals and has co-authored (with Paul Spirakis) a book on Probabilistic Techniques. He has been invited speaker in important international scientific events and Universities. He has been a referee for the Theoretical Computer Science (TCS) Journal and important international conferences (ESA, ICALP). He has participated in many EU funded R&D projects (ESPRIT/ALCOM-IT, ESPRIT/GEPPCOM). He currently participates in 6 Fifth Framework projects: ALCOM-FT, ASPIS, UNIVERSAL, EICSTES (IST), ARACNE, AMORE (IMPROVING). Jose Rolim is Full Professor at the Department of Computer Science of the University of Geneva where he leads the Theoretical Computer Science and Sensor Lab (TCSensor Lab). He received his Ph.D. degree in Computer Science at the University of California, Los Angeles working together with Prof. S. Greibach. He has published several articles on the areas of distributed systems, randomization and computational complexity and leads two major projects on the area of Power Aware Computing and Games and Complexity, financed by the Swiss National Science Foundation. Prof. Rolim participates in the editorial board of several journals and conferences and he is the Steering Committee Chair and General Chair of the IEEE Distributed Computing Conference in Sensor Systems.     

Modulation classification in the presence of adjacent channel interference using convolutional neural networks

This paper investigates a vital issue in wireless communication systems, which is the modulation classification. A proposed framework for modulation classification based on deep learning (DL) is presented in the presence of adjacent channel interference (ACI). This framework begins with the generation of constellation diagrams from the received data. These constellation diagrams are fed to convolutional neural networks (CNNs) for modulation classification. The objective of this process is to eliminate the manual feature extraction from the received data and make feature extraction process as a built‐in step with CNNs. Three types of CNNs are considered in this paper and compared for this objective. These types are AlexNet, VGG‐16, and VGG‐19. The proposed classifier is applied on Rayliegh and Rician fading channels.     

一种用于受训神经网络数据挖掘的遗传算法

将受训神经网络应用于分类领域时如何更好地抽取符号化规则是当今学术界广泛研究的问题。随着网络节点数和连接成几何级数增长,以前那种对网络连接和输出值进行全面分析的方法不再适用。提出了一种新颖的遗传算法用于从受训神经网络中提取符号化的规则。经实验证明这种方法对于提取规则是可行的。     

FPGA-based acceleration for binary neural networks in edge computing

As a core component in intelligent edge computing, deep neural networks (DNNs) will increasingly play a critically important role in addressing the intelligence-related issues in the industry domain, like smart factories and autonomous driving. Due to the requirement for a large amount of storage space and computing resources, DNNs are unfavorable for resource-constrained edge computing devices, especially for mobile terminals with scarce energy supply. Binarization of DNN has become a promising technology to achieve a high performance with low resource consumption in edge computing. Field-programmable gate array (FPGA)-based acceleration can further improve the computation efficiency to several times higher compared with the central processing unit (CPU) and graphics processing unit (GPU). This paper gives a brief overview of binary neural networks (BNNs) and the corresponding hardware accelerator designs on edge computing environments, and analyzes some significant studies in detail. The performances of some methods are evaluated through the experiment results, and the latest binarization technologies and hardware acceleration methods are tracked. We first give the background of designing BNNs and present the typical types of BNNs. The FPGA implementation technologies of BNNs are then reviewed. Detailed comparison with experimental evaluation on typical BNNs and their FPGA implementation is further conducted. Finally, certain interesting directions are also illustrated as future work.     

基于无线自组网技术的监控系统设计

设计一套基于无线自组网技术的监控系统,旨在对运输及库存中的重要产品进行远距离监控,避免繁琐的人工管理过程。从通信组网、硬件设计方面介绍了初步方案设计,拟利用短距离、低耗的WSN实现相对静止空间内的组网,利用MANET实现相对运动时的组网,以实现全国范围内的,信息传递时间小于5分钟的动态监控网络。     

TGSOM:一种用于数据聚类的动态自组织映射神经网络

针对传统Kohonen自组织特征映射(SOFM)神经网络模型结构需预先指定的限制,提出一种新的树形动态自组织映射(TGSOM)神经网络,当用于数据挖掘时该网络以其生成速度快可视性好具有显著优越性。该文详尽描述了该网络模型的生成算法,研究了算法中扩展因子的作用。扩展因子与训练样本数据的维数无关,其作用是控制网络的生长,扩展因子可以反映数据聚类的精度,即扩展因子值的大小与聚类精度的高低成正比。在聚类的不同阶段使用大小不等的扩展因子还可以实现层次聚类。     

基于数据挖掘技术的医院医保信息管理方法

为了解决医院医保信息管理中数据结构复杂、关系网难以建立、数据分析不全面、审核难度大等问题,文中设计了医院医保信息管理系统.首先,建立数据仓库将医院医保信息进行整合,其次,设计数据挖掘模块对医院医保信息进行处理,最后应用到医院医保信息管理中去.经实验表明,文中设计的挖掘模块中改进Apriori算法的数据处理速度是Apri...     

A NEW METHOD FOR SOLVING MSDE BASED ON WAVELET NEURAL NETWORKS

In this paper, a new method to solve multiscale difference equation(MSDE) with the M-band wavelet neural networks is proposed. It is shown that the method has many advantages over the existing methods and enlarges the range of the solvable equations.     

Hadoop下基于贝叶斯网络的气象数据挖掘研究

基于朴素贝叶斯的分类器是气象数据挖掘中比较传统的方法,但由于算法要求各属性相互独立,预测精度无法达到要求,且在处理海量数据时算法计算效率受到制约,对此提出一种Hadoop平台下基于离散贝叶斯网络的数据挖掘改进算法。算法不要求属性之间相互独立,且充分结合Hadoop平台适应处理大数据的优点,利用海量数据分析地面气象因素与温度之间的相关性,并由此选取预测因子来训练贝叶斯网络分类器模型,以达到预测温度的目的。实验结果表明,算法不但预测精度明显高于目前短期气候预测中采用的朴素贝叶斯算法,而且极大地提高了运算效率。     

NLDA non-linear regression model for preserving data privacy in wireless sensor networks

Recently, the application of Wireless Sensor Networks (WSNs) has been increasing rapidly. It requires privacy preserving data aggregation protocols to secure the data from compromises. Preserving privacy of the sensor data is a challenging task. This paper presents a non-linear regression-based data aggregation protocol for preserving privacy of the sensor data. The proposed protocol uses non-linear regression functions to represent the sensor data collected from the sensor nodes. Instead of sending the complete data to the cluster head, the sensor nodes only send the coefficients of the non-linear function. This will reduce the communication overhead of the network. The data aggregation is performed on the masked coefficients and the sink node is able to retrieve the approximated results over the aggregated data. The analysis of experiment results shows that the proposed protocol is able to minimize communication overhead, enhance data aggregation accuracy, and preserve data privacy.