Abstract: This paper presents a Bacterial particleswarm optimization (BPSO) strategy based on Particleswarm optimization (PSO) and Bacterial foraging algorithm (BFA). The velocity updating and position updating rules of particles are reinforced by two bacterial behaviors, i.e. reproduction and elimination-dispersal. Reproduction is applied to speed up the convergence rate,and elimination-dispersal provides diversity to overcomepremature problem and escape from local optima. Varioussimulation results have demonstrated that the performanceof BPSO is desirable in comparison with PSO and BFA.
Abstract: Segmentation of medical images is challenging due to poor image contrast and artifacts that result in missing or diffuse organ or tissue boundaries. In thispaper, we proposed a modification to the original level setalgorithm for implementation of deformable models. Themodification is derived from intensity averaging of the image. This new method helps to segment medical imagesaccurately into multi-level images. The level set algorithmhas some advantages over the classical snake deformablemodels but it has diffculties with large gaps in the boundaries of segmented regions. Such boundary gaps may causeinaccurate segmentation that requires manual correctionby users, while our goal was to keep user assistance toa minimum. The proposed method possesses an inherentproperty to detect gaps within the object with a single initial contour and also does not require specific initialization.
Abstract: Abstract — Comparing information secretly is the fundamental to secure multiparty computation. The knownprivate comparison protocols without a third party canonly resolve the Greater than (GT) problem and can onlycompare two integers. In this paper, we develop an improved cross products protocol by using the Paillier's additive homomorphic encryption at first. Then, we propose a three-round protocol for solving the private comparison problem in the semi-honest setting based on theproperty of the cross products. In comparison with theknown private comparison protocols, our method can determine whether a > b, a < b or a = b for two numbers a andb by running our proposed protocol only once. Moreover,our protocol can compare two real numbers in additionto integers. The analysis shows that our method is morepowerful than previous methods.
Abstract: Networks-on-chip (NoC) has attracted agreat attention to resolve the interconnection problem inbillion-transistor era. Network congestion has a negativeeffect on the performance of NoC. In this paper, we propose a novel congestion control method, adaptive traffcscheduling, for application-specific NoC. More precisely,given the traffc profile of the target application, the proposed algorithm automatically allocates the traffc acrossthe on-chip network, so that the network congestion is minimized. The traffc scheduling method can be implementedin both centralized and distributed manner for NoC. Thesimulation results show the performance of traffc scheduling is better than other congestion-control methods, suchas congestion-aware routing.
Abstract: In this paper, a SPICE sub-circuit modelis presented for 0.51m MOSFETs at liquid nitrogen temperature (77K). The voltage-dependent resistors rg (Vgs)and rd (Vds) are introduced to model the freeze-out effect,which can't be solved by BSIM3. The detailed parametersextraction technique has also been suggested. The modeling results show a good agreement with the experimentalresults, which proves the proposed model and parametersextraction technique are significative and can be applied inthe low temperature circuit design.
Abstract: This paper presents a Distributed datamining platform on Grid services pool (DDM-GSP), whichcombines grid services pool with distributed data miningto solve problems of traditional distributed data mining.Meanwhile, this paper implements parallel distributed genetic algorithm to resolve complex function optimizationon basis of DDM-GSP. Simulation experiments show thatfor concentrative mining, with the augmentation of population sizes, the convergent speed of standard genetic algorithm is improved by 39 times and computing time is improved by 81 times. However, improvement of populationdimensions, the average consumptive time of distributedgenetic algorithm on grid is about 31.7% less than standard concentrative genetic algorithm, while about 28.6%by contrast to traditional parallel genetic algorithm.
Abstract: Negative bias temperature instability(NBTI), which causes temporal performance degradationin digital circuits by affecting PMOS threshold voltage,has become the dominant circuit lifetime reliability factor.Design for lifetime reliability, especially for NBTI-inducedcircuit performance degradation, is emerging as one of themajor design concerns. In this paper, an NBTI-aware dualVth assignment is for the first time proposed to simultaneously reduce the circuit leakage current and ensure thecircuit lifetime requirement. Our experimental results onISCAS85 benchmark show that the NBTI-aware dual Vthassignment not only assigns more high Vth gates in the ISCAD85 circuits and leads to up to 14.88% (average 3.46%)further leakage saving under 5% circuit performance relaxation, but also brings different optimal high Vth (on average11mV higher) without performance relaxation.
Abstract: Semi-bent functions in even number ofvariables are a class of important Boolean functions whichhave three values of Walsh coeffcients, and so far only afew classes of such semi-bent functions have been found.In this paper, we give some concrete constructions of semibent Boolean functions in even number of variables.
Abstract: This paper proposes a new energy-savingrouting algorithm to improve the survival time of the ad hocnetworks as well as other performance metrics. The ESDSR algorithm introduces two parameters which are thecount of hops and the energy parameter into the routingcost function, and adjusts these two parameters according to the nodes' energy in the network by using a kindof weighted factor, thus achieving the purpose of adoptingdifferent routing strategies in accordance to the differentsituations of the network. Simulation results indicate thatcompared with DSR and MMBCR protocols, ES-DSR caneffectively improve the survival time, the throughput andother performance of ad hoc networks.
Abstract: For tree XML, constraints that specifystructural relationships among nodes or paths are very natural. In this paper, we introduce the concept of structuralintegrity constraints for XML (XSICs), which specify pathimplication, path cooccurrence, path mutual-exclusion, element obligatory inclusion and exclusive inclusion, anddefine the syntax and semantics of XSICs. For reasoning about XSICs, we rewrite all the other constraints intopath implication constraints, and develop a sound and complete set of inference rules for path implication constraints.Meanwhile, we propose the concept of path implication closure. By using the path implication closure, we prove thecompleteness of inference rules, and determine the implication decision about XSICs.
Abstract: In the semi-honest model, Lin and Tzengproposed an efficient solution to the millionaires’ problem based on homomorphic encryption. They reduced theproblem to the set intersection problem by encoding theprivate inputs in a special way (the encoding method allows an efficient protocol for the set intersection problem).This paper modifies Lin and Tzeng’s protocol by introducing an Oblivious third party, and the modification is fairand secure in the same model. Compared with some ofthe previous fair protocols, our protocol relaxes the cryptographic assumption or the trust model. Compared withother previous fair protocols, our protocol are more efficient. We also consider security against malicious adversaries, i.e., malicious participants (the twomillionaires) andthird party.
Abstract: Pervasive computing is characterized bythe integration with communication and digital mediatechnology embedded to the people's living space. People can transparently access the digital service anywhere.Wireless sensor networks are a novel technology and havebroad application prospects. With the maturity of thewireless sensor networks technology, pervasive computingis becoming a reality. It is become a new technology challenge to process the data streams of sensor networks forpervasive environment effciently and to find useful knowledge in these data streams. A k-means data stream clustering algorithm based on sensor networks is presented.The main idea of this algorithm is to select the initial centroids according to the aggregation gain of the node, thento cluster the data stream using the average square error.The experimental results are showed that this algorithm iseffective and effcient.
Abstract: In this paper, a new classifier of speakeridentification has been proposed, which is based onBiomimetic pattern recognition (BPR). Distinguishedfrom traditional speaker recognition methods, such asDWT, HMM, GMM, SVM and so on, the proposed classifier is constructed by some finite sub-space which is reasonable covering of the points in high dimensional spaceaccording to distributing characteristic of speech featurepoints. It has been used in the system of speaker identification. Experiment results show that better effect couldbe obtained especially with lesser samples. Furthermore,the proposed classifier employs a much simpler modelingstructure as compared to the GMM. In addition, the basicidea "cognition" of Biomimetic pattern recognition (BPR)results in no requirement of retraining the old system forenrolling new speakers.
Abstract: On the condition of electric-LO phononstrong coupling in unsymmetrical parabolic confinementpotential Quantum dot (QD), we obtain the eigenenergiesof the ground state and the first-excited state, the eigenfuctions of the ground state and the first-excited state byusing variational method of Pekar type. This system inQD may be employed as a two-level quantum systemqubit.The phonon spontaneous emission causes the decoherenceof the qubit. The relations between decoherence time withthe coupling strength, the confinement strengths in thexy-plane and the z direction, the coeffcient dispersion arediscussed.
Abstract: The problem of model selection for Support vector machines (SVM) with general Gaussian kernels is considered. Unlike the conventional standard singlescale Gaussian kernels, where all the basis functions have acommon kernel width, the general Gaussian kernels adoptsome linear transformations of input space such that notonly the scaling but also the rotation is adapted. We proposed a gradient-based method for learning the optimalgeneral Gaussian kernels by optimizing kernel polarization.This method can find a more powerful kernel for a givenclassification problem without designing any classifier. Experiments on both synthetic and real data sets demonstratethat tuning of the scaling and rotation of Gaussian kernelsusing our method can yield better generalization performance of support vector machines.
Abstract: This paper is concerned with the data association problem of distributed long baseline 2-D passivesensors (or jammed 3-D active sensors). Two data association methods, minimum miss distance method and hingeangle method, are discussed when the earth curvature isconsidered. According to the pioneering work, when theearth curvature effect was not considered, the results obtained by using the hinge angle based statistic were essentially the same as those obtained by using the miss distancebased statistic in two sensor case. In long baseline case, theearth curvature must be considered. In this paper, the detailed expressions of the above two methods are derived forlong baseline cases. In order to analyze the performance ofthe two methods, the correct association probability of thetrue targets and the incorrect association probability of theghost targets are defined. The Monte Carlo simulations intwo and three sensor cases are made, respectively. Simulation results show that the minimum miss distance methodis superior to the hinge angle method both in the case ofthree sensors and in the case of two sensors with targetsbeing located near the baseline, and that the performanceof the minimum miss distance method is equivalent to thatof the hinge angle method when the targets are located farfrom the baseline.
Abstract: In this paper, we present a 3D contentgeneration pipeline for filming, modeling and rendering adynamic real-world object. We construct a dome with widebaseline multi-camera setting to capture multi-view videoof motion actors. A new graph-cuts based 3D reconstruction algorithm is also proposed to achieve high reconstruction quality that can match the captured video. Finally, weintroduce a spatiotemporal interpolation technique, whichis suitable for our extremely wide baseline camera arraysto interpolate views at any viewpoint and any time. Experimental results show that our pipeline can model fastmoving scenes, which enable users to enjoy the 3D replaying in slow motion with especially high quality.
Abstract: The zero-crossings of image Laplacian isimportant for edge detection, which can accurately determine the edges of the image. In this paper, we propose anovel active contour model that utilizes the image Laplacian to construct an energy functional. We minimize thisfunctional and get a term which is related to typical image segmentation that the boundary is the zero-crossingsof image Laplacian. In order to improving the ability toresist noise and extending the capture range of the forcebased on this energy functional, we propose another energyfunctional of total variation for image Laplacian. Moreover, our model is incorporated with a variational levelset formulation without re-initialization proposed by Liet al. Therefore, re-initialization is unnecessary. In addition, interior contours are automatically detected withonly one initial contour. Comparisons with other majoredge-based or region-based models, such as Geodesic active contours (GAC) and the piecewise constant model (CV model), show advantages in segmentation of images withweak edges or intensity in-homogeneity.
Abstract: This paper proposes a cross-layer packetscheduling framework with QoS (Quality of service) mapping to provide UEP (Unequal error protection) at bothapplication and MAC (Medium access control) layers byinvestigating the CSMA/CA (Carrier sense multiple access with Collision avoidance) protocol from the queuetheory. We take full advantage of the rate-adjustabilityof shortened-and-punctured RS (ReedSolomon) codes, anddiscuss three essential aspects of the framework: the packetloss probability under both FEC (Forward error correction) and delay constraint, the mapping mechanism frompackets loss distortion to retransmission limit, and thecongestion-aware scheduling algorithm. Besides, a reservation factor for queue space is introduced to restrain thenegative effect of burst behavior of the parity packets. Simulation results indicate that the performance of our schemeis superior to the conventional methods.
Abstract: Since 3D models have been widely appliedin many research areas, the techniques for content-based3D model retrieval become necessary. In this paper, anovel visual based 3D shape descriptor called MATE isproposed. A modified Principal component analysis (PCA)method for model normalization is presented at first. Secondly, a new Adjacent angle distance Fourier (AADF) algorithm is proposed. Then the original two-viewed Dbuffermethod is presented to extract characteristics of projectedimages. Finally, based on the modified PCA method, theshape descriptor MATE is proposed by combining AADF,Tchebichef and two-viewed Dbuffer. Experimental resultsshow that the descriptor MATE provides better retrievalperformance than the best current descriptors.
Abstract: The present paper proposes a new cerbellar learning scheme of feedback error learning scheme fortracing in motor control system. In the scheme, the modelof cerebellar cortex is regarded as the feedforward controller. Specifically, a neural network and an estimator areadopted in the cerebellar cortex model which can predictthe future state and eliminate faults caused by time delay.Then the new scheme was used to control inverted pendulum. The simulation experimental results show that thenew scheme can learn to control the inverted pendulum fortracing successfully.
Abstract: This paper presents a physically-based virtual Chinese hairy brush model which can be used as adigital tool for learning or creating Chinese calligraphy orChinese painting. The whole framework of the proposedmethod comprises three components, i.e., brush geometrymodel, brush dynamic model and ink model. Instead ofsimulating the brush with bristles, our geometry modeluses points to simulate the whole brush bundle, which candrastically decrease the complexity inherent in the conventional bristle-level approach. The dynamic model describesthe behavior of the real brush's deformation in terms ofthe interaction of the external and internal forces with thevirtual writing paper, and a spring network is derived tocalculate the physical deflection of brush according to theforce exerted on it. The ink model describes the process ofink depositing in which brush geometry model is instancedand transformed into a real screen presentation.
Abstract: Task scheduling is one of key issues in gridcomputing. This paper focused on the task schedulingproblem with a large scale of independent and identicaltasks. An improved task scheduling algorithm DMIP is putforward, which is based on time and cost constrains combined with Integer programming but can control the maxnumber of tasks dynamically. Compared with plain Integerprogramming algorithm, named IP, which only consideringtime constrains, DMIP algorithm reduced both loss ratiosof tasks during the submitting process and the total execution cost by simulation experiments.
Abstract: As an important linguistic resource, collocation represents a significant relation between words.Automatic collocation extraction is very important formany natural language processing applications such asmachine translation, information extraction, and information retrieval. While traditional collocation extraction approaches are based on linguistic corpus, we propose to acquire collocations from the Web. Three classical lexical association measures (co-occurrence frequency, mutual information and t-test) are used to automatically extract collocation. Based on the experimental results, the benchmarksindicate that superior performance of this new Web-basedapproach in both high precision and recall.
Abstract: This paper proposes a fast and precise dequantization algorithm based on Chebyshev approximationpolynomial for MP3/AAC decoding. Due to the propertiesof Chebyshev approximation polynomial, compared withany other polynomial of the same degree, the infinity normof proposed algorithm reaches minimum. In contrast to theconventional method, the average error is reduced from0.03161 to 0.00051066 at the cost of two additional multiplications. The SNR (signal to noise ratio) for MP3/AACfiles is enhanced about 8dB averagely.
Abstract: It seems that all existing on-line/offline (threshold) signature schemes based on ShamirTauman's paradigm suffer from the key exposure problemof chameleon hashing. In this paper, we firstly proposea new effcient generic on-line/off-line threshold signaturescheme without key exposure. Compared with Crutchfieldet al.'s scheme, the advantages of our scheme are the lowercomputation and storage cost in off-line phase and lowercommunication cost in on-line phase. Meanwhile, we proveour scheme can achieve the desired security notions.
Abstract: Optical label switching (OLS) has beenregarded as an effcient technique to route and forwardIP packets transparently in the optical layer. In labelswitching system, Modified duobinary Return zero (MDRZ) label and Differential quadrature phase-shift-keying(DQPSK) payload have not been used yet. We proposea novel optical label switching scheme by the combination MD-RZ label and DQPSK payload for the first time.The transmitter is set up by cascading two Phase modulators (PMs) and one dual-arm LiNbO3 modulator. Dutycycle and extinction ratio of the MD-RZ labels are tunable. The DQPSK payload can be used in a 20Gbit/s optical label switching system, with 2-bit per symbol modulation/demodulation. The transmission performance of2.5Gbit/s MD-RZ label is also analysed. MD-RZ labelingof a DQPSK payload has been shown to be a promisingcandidate for the implementation of OLS system.
Abstract: Contour extraction is an important taskin image processing and computer vision. The contextualmodulation is a universal phenomenon in the primary visual cortex (V1). A biologically motivated computationalmodel is presented for contour extraction in this paper.Two mechanisms of contextual modulation, surround suppression and collinear facilitation, are integrated in thismodel. We obtain good results via this model to extractcontours from images with noise and texture backgrounds.This work provides a biologically motivated approach withgreat potential for computer vision.
Abstract: Correct classification of different metaboliccycle stages to identification cell cycle is significant in bothhuman development and clinical diagnostics. However, ithas no perfect method has been reached in classificationof metabolic cycle yet. This paper exploringly puts forward an automatic classification method of metabolic cyclebased on Biomimetic pattern recognition (BPR). As to thethree phases of yeast metabolic cycle, the correct classification rate reaches 90%, 100% and 100% respectively.
Abstract: This paper proposes an asymptoticallyoptimal watermark detector in Discrete cosine transform(DCT) domain images. The Gaussian mixture model asan analytical approximation of alpha-stable distribution ischosen to statistically model the Alternative current (AC)DCT coeffcients. Given the additive embedding modeland based on the theory of weak signal detection in nongaussian noise, the derivation of the detector is presented.The experimental results demonstrate the superiority ofthe new detector over correlation based detector and Generalized Gaussian distribution (GGD) based detector.
Abstract: The correlation characteristics of Multiview video (MVV) are influenced by the content of thevideo, illumination change, speed of moving objects andcameras, camera distance, frame rate, etc. In this paper, a framework of Multi-modal multi-view video coding(MMVC) is proposed on the basis of correlation analysisto achieve optimal performances among high compressioneffciency, low complexity, low memory cost, view scalability and fast random access. Different prediction modesare designed to fit MVV with different correlations andmeet different requirements of the Multi-view video coding (MVC). An optimal prediction mode is adaptively selected from the candidate modes according to the correlation characteristics of MVV. Experimental results haveproved that MMVC not only has best random accessibility, but also has outstanding performance in compressioneffciency, low memory requirement, low complexity andview scalability. MMVC is regarded as the most effcientand balanced MVC scheme among the compared schemes.
Abstract: This paper presents ND (Noise functiondetection) steganalysis algorithm to detect LSB steganography. The purpose of constructing the noise function isto quantify the smoothness or "regularity" of the images.ND method is based on the property that LSB embeddingwill increase the noise functional value of the image, andthen the LSB embedding message ratio is estimated byconstructing the simple line equation with the statistics ofnoise function in image. Experimental results show thatthis algorithm is more accurate and has a lower missingrate and false alarm rate than the conventional RS. Compared with RS method and some other powerful steganalysis approaches presented recently, ND method directly usethe noise function to estimate the LSB embedding messageratio, neither fixed nor dynamic mask is needed. Thus,more running time is saved. The ND method is relativelyfaster, simpler and has good detection result.
Abstract: We present a new public key trace and revoke scheme, which is suitable for scenarios where usersneed to join or leave the system frequently. The proposedscheme supports public traceability and dynamic registration and revocation. It is also fully collusion resistantagainst adaptive chosen ciphertext attack, and its provablesecurity is tightly reduced to the Decisional Diffe-Hellman(DDH) assumption in the standard model. No master keyis needed in our scheme, and it allows users to generatepublic/private keys themselves and then submit the public keys to the system. This means that the setup phaseonly involves an authentic channel instead of a secure one.Therefore, our scheme can take advantage of Public key infrastructure (PKI) with Certificate authority (CA) playingthe role of the system center.
Abstract: Multiparty key agreement (MKA) is a secure and robust approach to establish a group key forsecure group oriented applications over non-private underlying networks, and it can be effciently implementedwith ECC (Elliptic curve cryptography). In this paper, wepropose a generic two round model for ECC based MKA.All users involved are organized into a ring in this modeland distributedly compute a group key by two round message exchanges. Upon this model, we construct a concrete ECC based MKA protocol with scalability and lowcomputational complexity. The protocol is dynamic, thatis, it supports key update when user group membershipchanges to ensure forward secrecy and backward secrecy.The protocol is proved to be secure under the DecisionDiffe-Hellman (DDH) assumption on the elliptic curve.Its security is reduced to the intractable DDH problem onthe elliptic curve.
Abstract: In the paper, an opportunistic cooperationscheme is proposed to make an effcient use of the cooperative relay, where whether the cooperative transmissionwill be adopted depends on the channel quality from thesource to the relay, as opposed to the deterministic cooperation regardless of the relay channel conditions. Closedform expressions of the Bit error rate (BER) and the outage probability are derived over Rayleigh fading channelsto evaluate the performance of the proposed cooperativediversity scheme. For the purpose of comparison, the performance of the coded cooperation is also analyzed, withnumerical results justifying the merits of our opportunistic cooperation scheme. From the numerical results, theproposed opportunistic scheme outperforms the coded cooperation across the whole SNR range in terms of the BERand the outage probability and moreover, it is less sensitive to relay-destination channel conditions compared withthe coded cooperation, further showing the superiority ofour scheme.
Abstract: Broadcast authentication is a critical security service in wireless sensor networks. A protocol namedμTESLA has been proposed to provide effcient authentication service for such networks. However, when applied to applications such as time synchronization and firealarm in which broadcast messages are sent infrequently,μTESLA encounters problems of wasted key resources andslow message verification. This paper presents a newprotocol named GBA (Generalized broadcast authentication), for effcient broadcast authentication in these applications. GBA utilizes the one-way key chain mechanism ofμTESLA, but modifies the keys and time intervals association, and changes the key disclosure mechanism accordingto the message transmission model in these applications.The proposed technique can take full use of key resources,and shorten the message verification time to an acceptablelevel. The analysis and experiments show that GBA ismore effcient and practical than μTESLA in applicationswith various message transmission models.
Abstract: The success of Skype has inspired Peer-topeer (P2P) based solution for satisfactory real-time multimedia service over Internet. However, there still existsome limits such as incapability of interoperability incurredby proprietary protocol. To address the problem, we construct a novel SIP based Peer-to-peer overlay for Voiceover IP (VoIP) with preferable search performance. Themain contributions includes: (1) A novel algorithm waspresented for end-nodes to join the clusters with locationawareness; (2) Super-node (SN) overlay network was builtbased on the small world models, taking into account bothunderlying geographical distance and relationship proximity. Analysis results and preliminary simulation demonstrate that the proposed protocols could significantly decrease user searching delay, which is required by large P2PSIP network for its scalability.
Abstract: Membership revocation is a delicate issueof group signature. Verifier-local revocation (VLR) is areasonable resolution, especially for mobile environments.Based on the DDH and q-SDH assumption, we propose anew VLR group signature scheme with Backward unlinkability (BU). It has shortest signature size and smallestcomputation overhead among the previous BU-VLR groupsignature schemes. Then, we extend our scheme to gainnon-frameability.
Abstract: This paper proposes a novel optimal control algorithm for multimedia traffc in IEEE 802.11eWLAN (Wireless LAN) and describes its control processbased on Markov process. In this algorithm, an optimization policy is utilized to assign idle channel for traffcswith different priority. The simulation results indicate thatthe proposed algorithm decreases the overhead caused bycollision detection and retransmission mechanism. Moreover, both access effciency and overall performance areenhanced remarkably, and wireless resource can also beemployed more effectively in the novel algorithm comparedwith PCF (Point coordination function).