Current Issue

• 2018, No. 03　Published:25 March 2018 Last issue    Next issue
  Columns in this issue:

• Improving the quality of network message forwarding

  by customizing RISC-V compressed ISA

   

   

  L Qianru,WANG Yanpeng,CAO Zhuang,WEN Mei

  2018, 40(03): 381-387. doi:

  Abstract ( 212 )   PDF (767KB) ( 209 )   　　 Review attachment

  Instruction stream delivery and instruction Cache failure are two of the important reasons for processor energy dissipation. Programs based on loosely coupled RISC instruction sets exacerbate such energy consumption. Cachelimited network devices such as routers and switches suffer more performance degradation and power consumption due to instruction related processing. This paper focuses on network packet forwarding, which is one of the most important network functions. Thought analyzing instruction characteristics of network packet forwarding, we redefine the RV32 compressed instruction extension set based on the RISC- instruction set architecture and test it by Spike simulator. Experimental results show that the optimized compression rate is reduced to 70% and the dynamic instruction compression rate is 90%. Meanwhile, under the same Cache conditions, our RV32C set can reduce the instruction Cache failure by 30%~70% in comparison to the standard RISC- set.

   

  A novel high-resolution ADPLL for

  high-performance SOC application

  ZHAO Xin，HUANG Jinming,HUANG Yongqin,HU Xiangdong

  2018, 40(03): 388-393. doi:

  Abstract ( 164 )   PDF (1102KB) ( 222 )   　　 Review attachment

  Phase Locked Loop (PLL) is an essential part of high-performance SOCs that provide the chip with a system clock. This paper presents a novel All-Digital PhaseLocked Loop (ADPLL) structure for high-performance SOC applications and a novel high-resolution Time-to-Digital Converter (TDC)  improves the phase detection precision and reduces the TDC phase noise and improves the PLL jitter performance. In the nanometer process, the ADPLL system is implemented by using digital standard cells, which solves the bottleneck problems such as poor portability to new process, big area of passive devices, and poor antinoise ability in analog circuits. The system has the maximum frequency of 2.6GHz and the jitter performance less than 2 picoseconds.

   

  Design and implementation of a virtual

  operating system based on domestic platform

  ZHANG Yichao,WANG Xingyan,CHEN Zuoning,ZHANG Yufeng

  2018, 40(03): 394-404. doi:

  Abstract ( 130 )   PDF (1065KB) ( 205 )   　　 Review attachment

  HPC operating systems face unique challenges that cannot be tackled by traditional macro kernel operating systems. These challenges range from concurrency and high efficiency, system flexibility and fault tolerance, heterogeneity, I/O and memory bandwidth to low noise. We present a virtual operating system architecture, which combines a virtual machine manager with a lightweight kernel. The virtual machine manager supports both time and spacesharing virtual machines. We implement a system prototype Hypervk based on the domestic platform and preliminary experimental results verify the feasibility and efficiency of the system.

   

  A 65 nm CMOS compositiveps level narrow pulse driver

  XU Chaolong,LAI Mingche,LUO Zhang,XIANG Yang,PANG Zhengbin

  2018, 40(03): 405-410. doi:

  Abstract ( 143 )   PDF (969KB) ( 235 )   　　 Review attachment

  Depending on the breakthrough of singlechip optoelectrical integration and the advance of the optical pulse train generator technology, a new optical interconnect technology optical serializer/deserialier (SerDes) technology is proposed. Compared with traditional optical interconnect technology, optical SerDes technology is faster, lower power consumption and higher integration. The performance and integration requirement of driver circuit that drives the optical switch producing narrow optical pulse in a relative long cycle is more stringent. A 65nm CMOS integrated ps level narrow pulse driver used for optical SerDestransceiver is proposed. The driver drives optical switch producing optical pulse, which is as narrow as 13ps in SMIC 65nm CMOS library. The power voltage range is in 1.4~2.0 V, and the clock frequency is from several KHz up to 25 GHz.

   

  Two information entropybased seeding

  methods for 3D flow visualization

  HUANG Dongmei1，DU Yanling1,2，ZHANG Lüwen1

  2018, 40(03): 411-417. doi:

  Abstract ( 123 )   PDF (711KB) ( 198 )   　　 Review attachment

  Effective seeding method is the key to influence the streamline distribution and to understand the underlying properties of flow field. Based on the accurate description of flow field variation and important features, this paper proposes two information entropybased seeding methods to solve the wellknown occlusion and cluttering issue. The first greedy seeding method locates interesting areas through the calculation of entropy values. The greedy seeding method is highly sensitive to the important features. The second Monte Carlo seeding method generates random inputs based on a probability distribution, and then defines the influence areas of input grid points as a circle in 2D and a sphere in 3D. Comprehensive experiments on multiple datasets show that the greedy seeding method can capture the important features efficiently and the Monte Carlo seeding method shows significant ability to obtain global variation. Besides, the combination of both methods can get more optimal flow field visualization.

   

  H-PCPIR-V：An optimized PCPIR-V

  algorithm based on Huffman code

  WANG Botao，LI Ang，CHEN Yuemei，DENG Shizhuo，CHANG Bohan，WU Junxue

  2018, 40(03): 418-430. doi:

  Abstract ( 118 )   PDF (2065KB) ( 145 )   　　 Review attachment

  With the privacy issues drawing more and more concerns, privacy protection techniques based on Computational Private Information Retrieval (CPIR) allow a user to retrieve data from a service provider without revealing the users query information. For largescale applications, there exists a gap between privacy protection techniques and its feasibility. For the problem that the CPIR algorithm needs long computing time so as not to be suitable for largescale data privacy protection, this paper proposes a CPIR nearest neighbor privacy protection algorithm (HPCPIRV) based on Spark and Huffman code. The HPCPIRV algorithm partitions the spatial data into Voronoi diagrams according to the points of interest in the data preprocessing stage, and then utilizes the Huffman code to compress the candidate data in order to reduce bit computation operation. Spark parallel framework is used for query grid parallel computing in the server side. The experimental results show that the computational cost of HPCPIRV algorithm is about 30% lower than that of PCPIRV algorithm on the server side, the computational cost of client is about 10% lower, and the communication cost is about 40% lower.

   

  Information centric networking based on

  protocol oblivious forwarding technology:

  Design,implementation,and application

  WANG Qiang1,2,GE Junqiang1,2,CHANG Kun1,2,WANG Xiaodong1,2,TIAN Ye1,2

  2018, 40(03): 431-438. doi:

  Abstract ( 129 )   PDF (796KB) ( 192 )   　　 Review attachment

  The core idea of Software Defined Network (SDN) is to decouplethecontrol plane and the data plane in the network andprovide users with an open programmable interface by means of centralized control. Information Centric Networking (ICN) focuses on the content name and the namebased content routing. Protocol Oblivious Forwarding (POF) is a SDN forwarding technology that supports custom protocols.With the research and development of future network technology, it is possible to achieve ICN based on SDN technology. Therefore, we design and implement an ICN network scheme based on POF technology, and build up the ICN testbed.The multiplayer video conferencing application indicates the correctness and feasibility of the scheme, which can not only forward ICN data packets but also achieve realtime applications well.

   

  Robustness of interdependent networks based

  on different subnet cascade mechanisms

  FU Danlong1，ZHU Shuhua1，YUAN Zhifeng2，LIANG Zhuoqian1,2，DENG Yuan2

  2018, 40(03): 439-444. doi:

  Abstract ( 88 )   PDF (985KB) ( 207 )   　　 Review attachment

  Previous researches on cascading failure of interdependent networks are based on the same subnet cascade mechanism. Based on the previous studies, the loadcapacity model is used to propose a cascading model based on different subnet cascade mechanism. The impact of different attack strategies on the robustness of interdependent network is analyzed. By simulating interdependent ER random networks and interdependent scalefree networks, the effect of different attack strategies is compared, and the influence of topology elements on different tactics is found. In addition, by analyzing the results of different cascading failure mechanisms under the same attack strategy, it is discovered that the symmetry of the cascading failure mechanism greatly affects the robustness of interdependent network.

   

   

  A three-party password authentication key agreement

  schemes based on chaotic maps with user anonymity

  WANG Caifen,CHEN Li,LIU Chao,QIAO Hui,WANG Huan

  2018, 40(03): 445-455. doi:

  Abstract ( 137 )   PDF (505KB) ( 247 )   　　 Review attachment

  In the threeparty password authenticated key agreement based on chaotic map, by using week passwords, users can share the session key in order to avoid security threats in the authentication process of a public key infrastructure or storing longterm key. By analyzing the chaotic mapbased password authenticated key agreement protocols proposed by Lee, we find that the agreement cannot change the password. Besides, it can only be applied to the twoway communication between the user and the server. In order to improve this scheme, we propose two useranonymous threeparty password authentication key agreement protocols based on Chebyshev chaotic map: one is based on synchronized clocks, while the other is based on nonces. The protocol based on synchronized clocks has less traffic, while the protocol based on nonces is easier to implement. The advantage of the two protocols is that users selects only one simple password for mutual authentication and key negotiation. The server does not need to protect the user password table, which can avoid the passwordrelated attacks. In addition, in the process of mutual authentication, the user uses a temporary identity and hash function to achieve the user anonymity, while enhancing the security of the protocol and reducing the number of messages in the communication process as well. As a result, the efficiency of the agreement is improved, with perfect forward security. And its security is proven by BAN logic.

   

  A moving object localization algorithm in a

  multi-static sonar system with sensor location errors

  CHEN Weiwei,WANG Xin

  2018, 40(03): 456-463. doi:

  Abstract ( 93 )   PDF (752KB) ( 165 )   　　 Review attachment

  In a multistatic sonar localization system, the sonar positions often include random errors that can seriously affect the target's localization accuracy. To tackle this problem, a moving target location algorithm based on time sum and Doppler frequency measurements is proposed. Firstly, the proposed algorithm reorganizes the nonlinear equations based on time sum and Doppler frequency localization mechanism into pseudo linear equations about target position, velocity and intermediate variables, and the initial moving target position and velocity are obtained by using weighted leastsquares estimates. Secondly, the correlation between the position, velocity and the intermediate variables is used to estimate the estimation biases of target position and velocity. Finally, an error correction technique is used to correct the initial target position and velocity estimates. The effectiveness of the proposed algorithm is theoretically analyzed under small measurement noise postulation and the numerical verification is performed by using Monte Carlo simulation.

   

  A multi-candidate electronic voting protocol

  with RLWE homomorphic encryption

  LOU Yu，ZHU Gengming

  2018, 40(03): 472-480. doi:

  Abstract ( 155 )   PDF (593KB) ( 238 )   　　 Review attachment

  The use of security protocols to protect the privacy of voter and ensure the fair and effective voting is the basis of electronic voting applications, but the complexity of the security protocol is the biggest obstacle to electronic voting applications. A multicandidate electronic voting protocol based on RLWE homomorphic encryption algorithm,which can support multicandidates and satisfy the privacy protection of voters. Based on additive homomorphic properties of the RLWE homomorphic encryption algorithm, this protocol protects voters by counting cryptograph in the count stage, and uses the nature of the Chinese remainder theorem to improve the counting ability. This  protocol can support multicandidates to vote and eventually know the final vote number of each candidate,and set up publicity institutions to show every voting process in each steps for public verification.

  A new shape-from-shading algorithm

  based on Phong model

   

  ZHAO Zhongbin,ZHANG Zhiyi,XING Caiyan

  2018, 40(03): 481-486. doi:

  Abstract ( 132 )   PDF (669KB) ( 187 )   　　 Review attachment

  Focusing on the big error in traditional ShapeFromShading (SFS) algorithms for hybrid surfaces, a new SFS algorithm for a single image under hybrid surfaces based on perspective projection is proposed. In this paper, the Phong reflection model is used to describe the reflectance property of the surfaces. Assuming that the light source is located at the camera’s optical center, the image irradiance equation is established for this model. Then, the image irradiance equation is transformed into the HamiltonJacobi Partial Differential Equation (PDE) including the shape information of the surfaces. Using the highorder LaxFriedrichs (LLF) flux splitting scheme and fiveorder Weighted Essentially NonOscillatory (WENO) scheme approximates the viscosity solution of the PDE. Finally, the heights of threedimensional surface are obtained. Experimental results show that themaximum error and the mean error of 3D surface recovery of the new algorithm are significantly reduced, compared with the traditional algorithm.

   

  Image retrieval based on visual vocabulary

  tree fusing color wordbag feature

  ZHANG Nan1,HAN Xiaojun1,2

  2018, 40(03): 487-493. doi:

  Abstract ( 155 )   PDF (798KB) ( 170 )   　　 Review attachment

  In the traditional bag of word model, the SIFT features are extracted from the gray space of image, which cannot reflect the color information of image. To solve this problem, we propose to use a visual vocabulary tree vector that fuses the color feature to represent image contents. SIFT features are extracted and the vocabulary tree is built to obtain the SIFT features of images.The Kmeans method is used to cluster the HSV values of all images in the image library so as to obtain the representation vector of the color word bag based on the HSV space, there by avoiding the quantization error brought by the traditional color histogram method.The fusion of SIFT features and color word bag features completes the fusion of global and local features of the image.
  Finally, by calculating the similarities of the fusion features and sorting them from high to low,the image retrieval is completed. In order to validate the effectiveness of the proposed method, we choose Corel image database to analyze the performance of the algorithm, evaluate it from subjective evaluation and objective evaluation criteria, and compare it with the traditional method. The results show that,compared with the single feature method, the proposal improves the retrieval performance of feature fusion. The average retrieval precision and the recall ratio of the feature fusion method are all improved to some extent.

   

  A spectral feature matching algorithm based on maximum pool

  BAO Wenxia1,2,YU Guofen1，HU Gensheng1,YAN Shaomei1

  2018, 40(03): 494-499. doi:

  Abstract ( 109 )   PDF (931KB) ( 133 )   　　 Review attachment

  In order to improve the accuracy and robustness of image matching algorithms based on spectral feature, a spectral feature matching algorithm based on maximum pool is proposed. Firstly, we use the neighborhood information of image feature points to extract spectral features that are invariant to rotation and the linearity of brightness changes. Secondly, we use the feature points described by spectral features as nodes and the Euclidean distance between feature points as edge properties, and transform the image matching problem into graph matching problem. Finally, the maximum pool matching strategy is introduced to obtain the graph matching results. A large number of experimental results show that the proposed algorithm improves the accuracy and robustness of the spectral feature matching algorithm.

   

  Local representation based classification

  and its application in face recognition

  YIN Jun1,2,YANG Wankou3

  2018, 40(03): 500-506. doi:

  Abstract ( 12 )   PDF (537KB) ( 75 )   　　 Review attachment

  A noisedetection based total variation

   method for image denoising

  JI Zhong，ZHAO Shuo，WANG Jian，LIU Li

  2018, 40(03): 507-514. doi:

  Abstract ( 109 )   PDF (850KB) ( 281 )   　　 Review attachment

  An adaptive Total Variation (TV)denoisingmethod based on noise detectionis proposed to restore the natural images corrupted by both additive Gaussian noise and randomvalued impulse noise. The improved image restoration method uses a twostage iterative framework: identification of pixels corrupted by impulse noise and TV based image restoration. Firstly, the local statistical value, i.e. Rank Ordered Absolute Difference(ROAD) is used to adaptively detect the pixels corrupted by impulse noise that contains no valid information. Secondly, the L2TV method is used to restore noise images. The two steps are processed iteratively to obtain the denoisedimages. The introduction of impulse noise level parameters in the noise estimation process has the advantage of being able to detect the impulse noise more accurately. The L2TV denoising method can remove the Gaussian noise well. The combination of the two effectively solves the problem that the TV algorithm misjudges image impulse noise as edge and results in false edge. Compared withthe stateoftheart methods, the proposed imagedenoising method, named TVROAD algorithm, can achieve superior restoration results and preserve image detail features.

   

  Image recognition based on an overlap

   sparse group deep belief network

  TIAN Jin,CHEN Xiuhong,FU Junpeng,XU Derong

  2018, 40(03): 515-524. doi:

  Abstract ( 119 )   PDF (1357KB) ( 176 )   　　 Review attachment

  Most hidden neurons in Deep Belief Network (DBN) are noise variables, and have  group structure correlation. Group Sparse Deep Belief Network (GSDBN) constrains the implicit neuron variables via group Lasso so as to achieve variable group selection. However, the group sparse model not  only ignores the case that some features belong to multiple groups simultaneously, but also has the problem that hidden neurons are not sparse. In this paper, we propose Overlap Sparse Group Deep Belief Network (OSGDBN), which introduces the overlap group structure and makes the hidden neurons sparse, based on Group Sparse Deep Belief Network. In addition, we also explain the reason that the OSGDBN is sparser than GSDBN. The recognition results on MNIST, USPS, ETH80 and face datasets show that OSGDBN has a higher recognition rate.

   

  A hand gesture recognition method

   based on Hu-GLCM model

   

  LIU Hui，DAI Zhaokun，WANG Long

  2018, 40(03): 525-532. doi:

  Abstract ( 99 )   PDF (891KB) ( 162 )   　　 Review attachment

  Aiming at the shortcoming that a single gesture feature cannot describe the gesture well in hand gesture recognition, a hand gesture recognition method based on Hu moments and GrayLevel Cooccurrence Matrix (HuGLCM) is proposed. Firstly the skin color model is used to segment the captured image to get the gesture area. Secondly, the single connected contour of the gesture is extracted by mathematical morphology and polygon fitting. The improved HuGLCM method is used to extract the geometric shape and texture features of the gesture and establish a template database. Finally, the gesture image is identified and classified by the extended Canberra distance. The experimental results show that the improved algorithm has an average recognition rate of more than 95% on seven kinds of gestures, and the computing speed is fast, which can meet the realtime requirements.

   

  Review of the key techniques of

  geographic information retrieval

  WANG Zhibao1,XIA Hao2,WANG Chengbo2

  2018, 40(03): 533-543. doi:

  Abstract ( 133 )   PDF (764KB) ( 214 )   　　 Review attachment

  Most of the available digital information is associated with locations or places on the Earth. Although queries submitted to search engines often contain geographic information, the traditional information retrieval methods based on keywords are not effective, as they neglect the spatial relationship in the queries. Geographic information retrieval outputs results by spatial semantic match between queries and documents set, so it becomes a research hotspot in information retrieval and geographic information science. A framework of geographic information retrieval system is proposed, and the key techniques such as geographic knowledge base, geographic information extraction, geographic information retrieval model, hybrid index structure, and visualization of information retrieval are classified and surveyed according to this framework. Based on the comprehensive comparison and analysis of existed researches, some suggestions are presented for future research, and a lot of references are provided.

   

  A RCM algorithm based on feature weights

  ZHANG Peng，DAI Yueming，WU Dinghui

  2018, 40(03): 544. doi:

  Abstract ( 87 )   PDF (546KB) ( 133 )   　　 Review attachment

  The clustering criterion of the conventional Rough CMeans (RCM) algorithm is based on the hypothesis that the attributes involved in clustering are equally important. However, in the natural scenario clustering problems, the different attributes have different effects on the clustering results. To address this issue, we propose a weighted RCM by weighting clustering attributes. Specifically, in order to filter out discerning clustering attributes that have a crucial impact on the clustering results, the algorithm assigns different attributes to different attribute weights by introducing a weight matrix. The experimental results show that the proposed method is able to extract the attributes, which improves the clustering accuracy.