最新刊期

In the process of human beings knowing or understanding an image,they get the cognition result from their imaginal thinking actions.Simulating the human brain and viewing from the imaginal thinking,a method of image feature extraction and an algorithm of image retrieval were proposed.After mapping each image to a high-dimensional vector (each vector can be seen as a point in the high-dimensional feature space),calculating the distance function of a pair of points can indicate the relation of two images.The experimental results show that this image retrieval algorithm acts better than other algorithms in retrieval efficiency and speed.

Backscatter coefficient is one of the important acoustic parameters with velocity and attenuation.It is often used in diagnosis and material design.Compared with velocity and attenuation measurement,backscatter coefficient measurement is more complicated both for equipment and computation.Unified method for backscatter coefficient measurement remains unsolved both in home and abroad.In this work,the author applies the ultrasonic backscatter model and sets up the measurement equipment based on pulseecho mode for a kind of tissuemimicking material.Influences of both ultrasonic focus position and data length on backscatter coefficient measurement are studied by experiments.Results show that the focus position and data length have minor impact on the backscatter coefficient measurement.

The interpolation method of medical image is the key technology in domain of 3-d reconstruction of medical image.In this paper,a new interpolation algorithm based on order morphology is presented.The algorithm is implemented using percent dilation and erosion operations to resolve edge illegibility of the other interpolation algorithms like linear interpolation.The order morphological interpolation technique is compared to an existing shape based interpolation method and its advantages include superiority capability on handling various cases such as the branching and holes problem (appearance and disappearance of information) and more accurate volume estimation.By simulation,the arithmetic is proved an effective method,and the produced slice image is similar to original slice image.

For main problems of the existing gaze estimation methods which are that they restrict the head movement of users and the issue of individual calibration,this paper presents a gaze estimation method based on stereo vision.A five-spot model for the eye optical axis on the condition of two cameras and two light sources is set up.The model estimates the threedimensional orientation of the eye optical axis,which only needs one straightline intersection operation.On this basis,we forms a new gaze estimation method.This method allows natural head movement and minimizes the calibration procedure to only one time for a new individual.All aspects of this method meet the real-time requirement,which provides an effective solution for human-computer interaction gaze tracking system.

Multi-mode intra prediction in H.264 is computationally expensive.For efficient implementation to reduce computing pressure,according to computation intensity and data parallelism in intra prediction,a multi-group intra prediction stream algorithm suited for imagine stream architecture is proposed based on stream execution model.In addition,stripmining technology is applied into streaming implementation to design an optimized strip-mined multi-group intra prediction stream algorithm.Experimental results show that H.264 intra coder using optimized stream algorithm processes 1280×720 HD video at 45.9fps,satisfying the requirements of real-time video applications.

Attribute reduction is one of the most important notions in rough set theory.It has been proved that finding all reducts of a decision table is a NPhard problem.Based on Wu’s method,a novel approach to acquire all reducts is put forward from the perspective of algebraic equations in this paper.The efficiency of this novel method can be illustrated by experiments with UCI datasets and synthetic datasets.

Twostep strategy has become a fundamental method for underdetermined blind signal separation and the estimation of mixing matrix is a prerequisite for the source’s recovery.This paper proposes a twostep blind source extraction method for incompletely sparse case.The approach firstly estimates the basis vectors (columns of matrix) of partial sources using singlesource intervals and then extracts the corresponding sources with minimum interference;it doesn’t depend on other basis vector except the corresponding basis vector,so it avoids the essential condition the matrix is recoverable.Several simulation experimental results demonstrate its performance and practices.

To reduce the performance lossing induced by head-of-line (HoL) blocking in traditional wormhole router,and eliminate the underutilizing of buffers when virtual-channel router is handling variable-length packets in Network-on-chip(NoC),in this paper,an on-chip wormhole router architecture with Dynamically Allocated Input-Queues(DAIQ) is proposed based on virtual channel technology.The proposed architecture allocates buffer resources dynamically to regulate the depth and number of virtual queues.Additional,a novel switch allocation mechanism-SRRM is proposed to keep granting flits from the same packet,and SRRM improves the latency and throughput of switch under high network load.The results of simulation and synthesis show that DAIQ router provides similar performance as a traditional on-chip virtual-channel router,but achieves about 50% savings in buffering resources and yields 30.18% savings of standard cells area and 38.4% savings of power under 013um CMOS process.

Considering the aggregation property of the rules which a packet matched,we combined counting bloom filter and hash table to design and implement an efficient multidimensional packet classification algorithm called CBHT(Counting Bloom filter and Hash Table).Based on the aggregation property,for fivedimensional packet classification,we first got the smallscale rule set which matching the IP address using counting bloom filter.In this limited rule set we processed search on the other three dimensions.CBHT improves the speed of packet matching effectively and supports rule set’s dynamic update.The experimental results show that CBHT algorithm saves 60% of the hardware resources than B2PC algorithm and the number of average memory accesses lower than B2PC 22.8%.

The noise of capacitive MEMS inertial sensor system can be depressed by increasing the mass of the seismic and initial sensing capacitance,which can not be obtained by deep ion etch process when the aspect ratio of comb capacitance is very large.Accordingly,a novel super high precision MEMS inertial sensor with the actuators drived by Lorentz force for increasing the initial sensing capacitance is developed in this paper.The initial sensitive capacitance is increased largely by the actuators,the mechanical noise is reduced to 0.61ug per square root Hz by etching damping slots in the sensitive combs of actuators.The resonant frequency obtained by ANSYS is 598Hz,and the static displacement sensitivity is 0.7μm per g acceleration of gravity.The grid capacitive inertial sensor is fabricated by MEMS silicon-glass bonding process,and the quality factor tested by which the sensor is drived by Lorentz force is 715 in air,which proves the feasibility of the idea that the novel MEMS sensor can be fabricated by siliconglass bonding process and its initial sensing capacitance can be increased by Lorentz force actuators.

QGA(Quantum Genetic Algorithm) is an intelligent algorithm which colligates the advantages of quantum computation and GA(Genetic Algorithm),which is often used to solve combinatorial problems.In this paper,the method of best polarities of multioutput RM (ReedMuller) logic circuits is given,and QGA is applied to the optimization of multioutput fixedpolarity RM logic,circuits.To deal with the defects of the easily immerging in partial minimum frequently,this paper proposes a QGA based multioutput RM best polarity search algorithm which combined with community disaster.Finally,through several largescale PLA format benchmarks testing,results show that QGA based search algorithm has higher performance than GA based in optimization,search and convergence.

A new method is proposed to identify complex dynamical networks,in which only some state viariables can be measured.For multioutput systems where not all state variables are measurable,a reduced-order observer is designed.The approach based solely on output variables to identify networks is further realized in this paper.A few conditions serving the observer are specified,derived from the Lyapunov Stability Theorem.The proposed method for designing observers is more convenient and practical than traditional ones.Its effectiveness is demonstrated by the simulation of a complex network given in the paper.

Real-time and robust image registration is the premise and key technology of aerial video stabilization,panorama stitching and ground moving target detection and tracking.This paper presents a novel scene complexity and invariant feature based aerial video registration algorithm.The main characteristics of the proposed method include:(1) Based on analyzing the key difficulties and challenges of aerial video registration,several new methods are presented to realize fast and effective video registration under various real scenes,include integral image based fast image scale space generation,scene complexity based feature number controlling,and statistical error distribution of correspond features based cascade filtering.(2) Through combining the multiscale Harris corner detection,SIFT (Scale Invarianu Feature Transform) feature description,and the RANSAC (Random Sample Consensus) based frame geometry transformation parameters estimation,the proposed algorithm achieves satisfied rotation,scaling,brightness invariance and accuracy of registration.Experiment results show that the proposed algorithm carries out realtime and precise image registration under complex conditions with change of scene,large image translation,scaling and arbitrary rotation,and the average processing speed for a resolution of 320x240 unmanned aerial video sequences achieves 20.7fps.

A new rate model based on Generalized Laplacian Distribution is introduced in this paper. Compared to the traditional Laplacian Distribution,the model can better describe the kurtosis and tail distribution of DCT of video sequences.Besides,according to the new rate model,a lowdelay rate control algorithm adopting target bit allocation and Quantization Parameter (Qp) on Macroblock (MB) level is proposed.The compute complexity,which is caused by Qp obtained from MB variance in TMN8,can be reduced in the algorithm by adopting the method that the basic Qp is determined in frame lever and adjusted in MB level.Simulation results show that the proposed algorithm can effectively regulate buffer occupancy and reduce system delay.

The existence of specular highlights is a great obstacle of shape-from-Shading.For a single gray-scale image with only intensity information,the existed highlights detection methods based on chroma or polarization analysis can not directly be applied on it.So,a method using surface shape is provided,it makes full use of the imaging process.Through the surface normal estimation,compute the maximum a posteriori probability of each reflection composition under simulated annealing and detect the highlights areas,finally remove the highlights based on the assumption of curvature continuity.Experiments reveal that this method is effective on synthetic and real-world images,improves the accuracy of surface recovery for image combined specular highlights.

A new security protocol automatic generation method has been proposed,which goes for fair exchange protocol.Due to the properties of fair exchange protocol design space,we got a fairness decision model.We introduce the channel code into the protocol string,and separate the own sets from belief set of the principals to implement the derivation of messages and complete the decision of goals.There are two kinds of goal of fairness,which are respectively local goals and global goals.Using fitness function genetic algorithm,the protocol space coding with binary is proceeding the optimize search and get the protocol met goals finally.At last,we give an example to explain the feasibility of our method.

By using Bloom filter,a multi-pattern matching engine suitable for hardware implementation was designed.The suspicious strings are picked up by the Bloom filter engine,and then sent to a bitsplit state machine for verification.As a solution for speedmismatching,a pipeline structure was purposed,which enabled the filter engine and the state machine to process strings in parallel to achieve a higher throughput.In addition,by using some optimal methods such as threshold for rules,the resource consumption is reduced and the scalability is improved while preserving the desired speed of original scheme.

CHOKe is an algorithm of fair approximation of the stateless and active queue management.The CHOKe hit,whose effectiveness and the force of punishing non-response flows are the crucial factors to enhance the fairness algorithm,is used to approximately identify and punish non-response flows.In this paper a new algorithm,S-CHOKe,which enhances the fairness of the CHOKe algorithm,not only replace the CHOKe hit with sampling CHOKe hit to improve the effectiveness of CHOKe hit,but also use Queue hit to adaptively determine the number of packet loss so to moderately punish non-response flows.Simulation results show that S-CHOKe,which is able to adapt to changes in flow quantity,is effective and fair.

Attestation is one of the critical problems to Trusted Computing.The solution project of attestation in TCG,due to bad extensibility,inflexibility,bad performance and leaking privacy,is becoming bottleneck to application,popularizing of Trusted Computing.In this paper,the formal concept of attestation is defined,functional structure and execution mechanism of the ternary and the quadruple attestation system are presented,and pointed out that the attestation of identity is a “push” quadruple attestation system,and the attestation of platform is still ternary attestation system,Moreover,the development about Attestation of Identity,Attestation of Platform and Attestation of Runtime Environment,and so on,are analyzed.This paper also presents a summary of the current state of these techniques,a discussion on the future research topics,and the challenges of attestation of TCG.

The real-world environment oriented Q&A (Question and Answering) system has gained general attentions recent years.A new question analyzing method—QICA (Question Information Chunk Annotation) was presented for complex questions.QICA classifies question information into five types and maps a question from plain text into a semantic structure.An M3Ns (Max-Margin Markov Networks) was used for QICA.The QICA results were used for question paraphrase recognition and the results prove its effectiveness.

In our approach,the global visual vocabulary which is similar to keypoints of codebooks is built with Gaussian Mixture Models based on local image features.Images are represe
nted as a new set of feature vectors which are summed posteriori responsibility
relative to different visual words.The discriminative classifier is trained by Support Vector Machine with linear kernels based on above features.Experiments were performed on the PASCAL VOC 2006 dataset and the results suggested the influence of background factors on classification effectiveness.And further experiments showed that the features extracted from object areas can be combined effectively to improve classification performance in our method.

A novel approach is proposed to improve low-dose CT image quality by a non-local weighs prior which is acquired from previous normal-dose scan.First,we align two images—the low-dose and the previous normal-dose scan image,and filter the low-dose image to suppress noise.Then,the weight matrix of the registered normal-dose image is calculated by the non-local means weighting formula,and the low-dose image is restorated using this weight matrix as a prior.Simulated phantom and perfusion data experiment results demonstrate a highly quality improvement in terms of suppressing noise-induced streak artifacts and preserving resolution of the low-dose image using the proposed method.The proposed method is especially effective for clinical short-term repeated CT scan,such as perfusion and radiation therapy.

A multi-wavelet system can simultaneously provide several features,such as orthogonality,finite support,symmetry and so on.PCNN that possesses biology vision background is the third generation neural network,so it has innate superiority used for image processing.Comprehensively utilizing the advantages of multi-wavelet and PCNN,this paper presents a image fusion algorithm based on match measure of pulse number-MMPN.In this algorithm,the linking strength of each nerve cell adaptively confirms value according to image contrast defined in multi-wavelet domain that is consistent with human visual characteristics.The experimental results show that the algorithm can effectively improve the entropy,standard deviation,and quality measure of the fused image,and has good performance.

Because of the uncertainty of outer circumstances and the complexity of human being′s face,how to track and depict the facial expression is a difficult problem.Based on it,the author puts forward a rather simple solution than such traditional methods as model recognition and sample learning.On the condition of the video capture,the author analyzes frame images,compares several edge detection methods,and uses a modeling method via Edge Feature Fetching for facial expression,in order to fetch face feature values and build a model to describe the facial expression.In addition,the author uses other methods,such as curve fit and model control,to describe the cartoon image of human being′s face and simulate the animation of 2D facial expression,and to complete developing the cartoon model drawing and describing realistically the change of facial expression.

With the proliferation of MP3 music,compressed-domain music information retrieval from the Internet has come into being an important and urgent research field.In this paper,we propose a novel compressed-domain audio fingerprinting algorithm based on MDCT spectral entropy.The input MP3 music file is first partially decompressed to obtain MDCT coefficients as intermediate results，whereby we calculate the MDCT spectral entropy through consecutive long windows and come to the final fingerprint sequence by magnitude relationship modeling.Such fingerprint exhibits strong robustness against various frequency-and time-domain audio distortions due to its statistically stable nature.Experimental results show that in our test database which is composed of 100 distinct Chinese pop songs，a 5s music clip is sufficient to identify its original recording in real time，with more than 90% top one precision rate even under various severe audio signal distortions.

Traditional network measurement methods,such as uniform sampling and poisson sampling,deviate true data set a lot due to the self-similarity of the network traffic.This paper proposes a new network data measurement and statistics model based on realtime analysis (NMSM) which can be used to analyze the realt-ime changing characters of the data,adjust the sampling frequency dynamically,then fulfill the data sampling to the best advantage.The experiment showed that if the network environment changed smoothly the NMSM model seldom affected network;and if the network environment changed greatly it fitted the true data set very well.

This paper describes the experiment on ionization radiation effects of device performance including the increase of dark current signal and the voltage shift of output amplifier of charge coupled devices (CCDs) by Cobalt-60[WTBZ]γ[WTB1] source under different ioni
zing dose rate and bias cases.It is analyzed about the ionization damage sensitivity of image sensor elements and output amplifier section.These studies have laid a good foundation for the development of standard experiment process and assessment method on ionization radiation effects of charge-coupled devices.

Steel level measurement in continuous casting tundish has been trapped in a problem:atrocious measuring circumstance caused by high-temperature of molten steel and the flux slag on upper layer make it difficult to measure molten steel level accurately.According to this fact,a new method based on temperature field information is put forward to molten steel level measurement.By means of analyzing temperature distribution of different layers in continuous casting tundish,including the air layer,the slag layer and the molten steel layer,air-slag layer and slag-steel layer interfaces location are located,and the molten steel level is measured accurately finally.Application of this method in field shows measuring error is 3.2mm,the maximal error is 5.2mm,and the minimal error comes to 0.3mm.It indicating this method is authentic and it has a favorable application prospects.

A practical dual-purpose method is presented not only to enhance heat transfer but also to prevent notching effect in deep reactive ion etching process,especially for a structure suspended by slim and long beams in overetching.A metal layer is sputtered on the bottom surface of top silicon to dissipate heat,therefore lower the temperature over the silicon structures.FEM simulation and experimental investigation are carried out to verify the effectiveness,respectively.Meanwhile,the metal layer prevents the charges of ionized radicals from accumulation on the dielectric layer of SOI or SOG,thus suppresses notching effect which is also validated by optical measurement.The applicability of the proposed method is examined by fabricating a SOI comb-finger actuator.

Major challenges in current computeraided diagnostic (CADe) schemes for nodule detection in chest radiographs (CXRs) are to improve the sensitivity and reduce the false positive rate.This paper proposed a new CADe scheme.First,an active shape model based algorithm was applied to segment the lung field.Then,seventy nodule candidates were selected and 27 features for each candidate were extracted.Linear classifier was employed for classification.Because of two-step nodule enhancement,only few nodules were missed at the candidate selection step.For feature extraction,watershed algorithm was applied to segment the nodule candidate.Based on the segmentation result,shapebased features,gray level statistics features, surface-based features and gradient features were extracted.Another feature based on Canny edge detector was designed to eliminate bone crossings as false positives.A publicly available database containing 126 nodules in 126 CXRs was used for testing our CADe scheme.A sensitivity of 72.2% with 4.5 FPs per radiograph was achieved in a leave-one-out cross-validation test.For the very subtle and extremely subtle nodules,a sensitivity of 52.7% was achieved at the same FPs.

In this paper,the stabilized biconjugate gradient fast Fourier transform method combined with a BP neural network technology is applied to the electromagnetic inverse scattering problem of composite metallic and material target．The first,the target scattered electric field is measured at some points by means of the BCGs-FFT，then the electric field scattered by the target are fed into the BP network，whose output are the electromagnetic parameter.After proper training，the inverse scattering model of target has been setup and the geometric and/or electromagnetic properties of composite metallic and material target are reconstructed in real-time．Numerical results are provided for the validation of the proposed approach．

Historically,nonplayer characters(NPCs) were simulated as part of the game logic.Unlike traditional techniques,a new intelligent controller (IC) was designed for autonomous NPCs(ANPCs),which separates AI logic from game engine.Users communicate with the IC by three interfaces which define sensing,thinking and acting of ANPCs.An extension to situation calculus is proposed to help implement the IC.The extension includes the introduction of epistemic fluent and the definition of linguistic epistemic fluent to solve uncertain effects of actions and the potentially uncountable number of possible worlds,and incorporating sensing into the effect axioms to make ANPC come back the real-world model from its cognitive model.A game AI platform based on the model is built.It proves to be especially convenient for creating ANPCs which can counteract with player characters by applying the platform to game Quake 2.

Most existing retinal blood vessels segmentation methods are robust only for normal retinal images,but not for pathological retinal images.In this paper,a new method for segmenting blood vessels in pathological retinal images is proposed.Firstly,the divergence of the vector field is used to locate most centerlines of pathological retinal image.Then the directional information of each pixel in centerlines is computed and the pixels around the centerlines are detected by modified directional local contrast method.Finally,the whole blood vessel network is obtained via reverse tracing at the end of each blood vessel segment.The proposed method has been tested with all the pathological retinal images in the publicly available STARE database.Experiment results show that the proposed method achieves an area under the receiver operating characteristic curve of 09426 and accuracy of 09502,it is distinctly better than methods of Hoover and Benson et al.Moreover,the proposed method has overcome the limitation of method of Benson et al,and is robust for all kinds of pathological retinal images.