摘要:In this paper,we discuss exponential stability of equilibrium point for asymmetric Hopfield neural networks with time delay via the method of constructing suitable Lyapunov functional,and obtain sufficient conditions for global exponential stability of equilibrium point.We also estimate the exponential convergence rate of the equilibrium point of the time delay neural networks.
摘要:This paper discusses a fault tolerant design and reliability evaluation method for artificial neural network(ANN) controller in detail,taking the Artificial Neural Network Fast Valving Nonlinear Adaptation Controller applied in power control system as an example.Firstly,the importance of the ANN reliability study and the generality of the fault tolerant design and reliability evaluation presented in this paper are proposed.Next,the design structure and principle of fault tolerant design of ANN controller are introduced.Finally,the formula of reliability evaluation and its extension are developed based on the result of fault tolerant design.
关键词:Artificial neural network controller;Fault tolerant design;reliability evaluation;Power system
摘要:Many new challenges have been emerging in the information security area such as encryption,digital signature,identification,etc..Public key cryptosystem is the only solution.Modular exponentiation of large operands is the kernel operation in many public key cryptosystems.It is also the bottleneck of the computing efficiency.Based on Montgomery algorithm,a novel systolic array for modular multiplication is presented in this article.Moreover,with the binary method,a 256 bit systolic modular exponentiation processor,THM256 has been successfully designed and fabricated with 0.8μm CMOS technology.The die area is 17 63mm 2.with 18677 equivalent gates.It can run at 90MHz or higher frequency under 5V and less than 1 5W of power consumption,which means the throughput of 117kbits/s for modular exponentiation calculation.
关键词:PSA public key cryptosystem;Modular exponentiation;Modular multiplication;systolic array
摘要:A new camera calibration technique for 3D machine vision with the radial alignment constraint (RAC) is described.This method seperates camera parameters by using RAC,gives a new camera model with one term radial lens distortion,and provides the solution through a linear algorithm according to the reasonable order instead of traditional nonlinear optimization,so it shows high accuracy and simplicity.The technique has successfully been applied to calibrate the CCD camera fixed on ZEISS CMM.
摘要:Unwrapping a 2 D inconsistent principal value phase field of PMP(Phase Measuring Profilometry),that may contain regions of discontinuous phase,is an ill posed problem.Based on the globally nonstationary and locally coherent property of the fringe image texture,this paper measures the instantaneous frequencies by multiple channel wavelet like Gabor filters.By comparing the channel amplitude and phase responses,boundaries and discontinuities in texture can be detected,thus the phase field of PMP is segmented to consistent regions.The technique is demonstrated by application to synthetic and real PMP image of body models.
摘要:In the filtering algorithm of Ref.,it is by directly comparing the absolute value of Corr 2( m,n ) after being rescaled and W(m,n) to determine whether the data point W(m,n) should be kept.And the standard deviation σ m of noise at each scale m is employed to stop the iterating.But in application simply doing like this will lead to a rough reconstructed signal.In this paper,the concepts of distinct coefficient COF and threshold coefficient TH are proposed to overcome the deficiency,and the step of edge extraction is also changed to make the choice of COF and TH conveniently.The edges of signal can be kept well while the noise is removed greatly by appropriately choosing COF and TH.The simulation shows that the filtering efficiency of the new algorithm is better than the original one.
Xu Qinwei, Li Zhengfan, Mao Junfa and Zhu Jianzheng
Issue 2, (1999)
摘要:A new macromodel of lossy transmission lines is presented for the time domain simulation of the interconnection systems in the high speed VLSI.It is based on the method of characteristics(MC) with some modification.By approximating the Telegrapher’s equations in the frequency domain,the time domain macromodel is obtained and a set of recursive formulas is derived for computing the voltages and currents at two ends of the lines.The algorithm is useful for both delay and crosstalk estimation and can be used to evaluate transient responses of high speed interconnection circuits.The new macromodel concerns only the quantities at two ends of the lines,so the algorithm based on it consumes less computer time and less memory in comparison with other methods.It is of high accuracy and absolute stability.
关键词:high speed VLSI;interconnect;Method of characteristics;Macromodel of lossy transmission lines;transient response
Lu Fuyun, Yuan Shuzhong, Pan Jiaqi, Gai Qi, Zhao Yuanchao, He Qingguo
Issue 2, (1999)
摘要:The fundamental principle to measure S parameter of microwave monolithic integrated circuit by using ps photoconductive sampling technique has been reported.The S parameter of a wide band (18~36GHz)low noise three stage amplifier has been measured using this method.The results agree well with those measured by using network analyzer,but 100GHz band width has been achieved by ps photoconductive sampling technique.
摘要:Considering the fact that test response observation and test application only use a fraction of test time,a test subsession partitioning scheme is presented.Some further sources of the test scheduling problem can be used.Subcircuits in conflict in the process of test are only partially in conflict now.A new test scheduling algorthm is proposed after the test subsession partitioning scheme has been combined.The effectiveness to search the test scheduling solution is improved greatly by using the previous conflict information.
摘要:A novel wavelet based fuzzy neural network is proposed in this paper,and its structure based on the fuzzied wavelet neural network and the related learning algorithm is presented too.It is shown that the wavelet neural network has good covergence properties and approximation ability in computer simulations.
摘要:This paper introduces the method that uses neural networks to realize Hough Transformation (HT) and deduces the relation of HT and Fourier Transformation (FT).The HT can be realized by Fourier optical system,so it solves the problem that HT requires much more computation and larger memory space.
摘要:Based on the method of “High Frequency Network Averaging”,an unified approach to modeling DC DC PWM converters in discontinuous inductor current mode operation is proposed.Low frequency behavior models are derived.These models,on the one hand,serve as a basis to derive differential equations of the converters,on the other hand,are very suitable for SPICE simulation,where DC and small signal characteristics can be easily obtained.Simulation and experimental results verifying the modeling are reported.
关键词:switching power converters;Discontinuous inductor current mode;modeling;Low frequency characteristics
Cheng Weisheng, Lu Jianhua, Zheng Junli and Zhang Yu
Issue 2, (1999)
摘要:The Flooding Routing Algorithm (FRA) is recommended by Eurocom.Due to its robustness,FRA has now been widely applied to packet switching networks.However,the inherent drawback of FRA is that the flooding broadcasting characters may inevitably cause excessive network traffic redundancy with degraded network performance.This paper develops a so called Direct Routing Algorithm coupled with a Rerouting Strategy.Theoretical analysis and computer simulations show that the proposed scheme can substantially improve the network performance while maintaining robustness.
摘要:In terms of the technology of photocurrent negative feedback,the squeeze of closed loop photocurrent fluctuation with -13dB,-14 8dB and -8 9dB squeezing in semiconductor devices (LED/LD) and tunable dye laser has been respectively generated and observed.These results have broken through the theoretical limit given by quantum demolition measurement,and can not be explained by the traditional quantum measurement theory.We employ the concept on anticorrelation state of photocurrent or light field,and correctly explain the above experimental phenomenon.
关键词:Photoelectric negative feedback;Squeeze of photocurrent noise;Quantum demolition measurement;Anticorrelation state
摘要:The effects of group velocity dispersion(GVD) on nonlinear optical loop mirror (NOLM) OTDM demultiplexer are analyzed theoretically and numerically in this paper.First the nonlinear Schrodinger equations which describe the interplay of GVD and nonlinearity are given.Then the split step Fourier method is used to solve the equations.The numerical results reveal that the soliton number,the dispersion sign,the width of pulses and the walk off time between the control and signal pulses must satisfy certain values to maintain the performance of the demultiplexer.
摘要:In this paper,the defect and yield computational models associated with IC lithography process have been built by Monte Carlo statistical method.A realizable system of IC functional yield simulator,XD YES,and an approach of yield design are obtained.The applications of XD YES have shown that the calculated results agree with the practical ones.
关键词:VLSI;functional yield;Prediction and simulation
摘要:This paper studys the application of Symmetric Code theory to optimization problem with constraints after the analysis of dynamic system.A dynamic optimal control problem is considered as an application.With regard to a system with no zero initial and final states conditions,General Horizontal Symmetric Code and General Vertical Symmetric Code theory is given out.After discussing one input and bi input linear system dynamic model,we propose Theorem 1,Theorem 2 and Theorem 3 to describe the property of Symmetric Code,which can automatically satisfy the final states constraints of a dynamic system.This theory offers a foundation to expend the applications of Symmetric Code theory.The initial application of Symmetric Code is of much better performance than classical Genetic Algorithms.We can also find its applications in machine learning and neural network technique.
Wen Guangjun, Li Jiayin, Liang Zhen, Xong Xiangzhen, Yang, Ziqiang
Issue 2, (1999)
摘要:A high power relativistic backward wave oscillator has been constructed and studied.A typical electron beam of 1 8kA is accelerated across a diode potential of 450 keV and then guided through a section of corrugated waveguide by an axial magnetic field of 20KG.Peak microwave power of 100MW at 9 3GHz has been observed in the circular TM 01 mode.We analyze the nature and frequency of the beam wave interaction of the BWO by use of consistent linearized theory,and describe the experiments and measurements to determine the frequency and power of the device.An eletromagnetic particle in cell code MAGIC has been used to investigate the nonlinear beam wave interaction.Comparisons between the experimentally measured results,and the analytic and numerical simulations of the device are made.
关键词:relativistic backward wave oscillator;high power microwave;Particle simulation
摘要:Software reuse offers a solution to eliminate repeated work and improve efficiency and quality in the software development.In the recent ten years,object oriented technology has appeared and become a mainstream technology,thereby providing fundamental technology support for software reuse.Software reuse regains more attention in software engineering research and is considered a pratical and feasible approach to solving the software crisis.Software reuse is generally classified into two catalogues:product reuse and process reuse.Reuse based on software components is the important form of product reuse and is the major area of software reuse research.At the same time,software component technology plays an important role in distributed object research.Therefore,software component technology is regarded as a key factor of successful software reuse.The development and application of software reuse technology will facilitate the revolution of software development and reorganize software industry.As a result,the development of software components will become an independent and inseparable industry.The revolution offers a good chance for Chinese software development.This paper is a summarization on the development of software reuse technology.It presents fundamental concepts and key techniques of software reuse.After introducing several successful research and practice in software reuse,including Jade Bird Project,a Chinese national key project supported by the government,it proposes some ideas on how to reinforce research and application of related techniques and facilitate the development of software industry in China.
摘要:The paper presents a survey of recent advances in deep submicron MOSFET’s hot carrier reliability.The main phenomena and physical mechanisms of the hot carrier induced MOSFET’s degradation are introduced,with emphasis both on the effect of hot electrons and on that of hot holes,which may lay the foundation for the research of deep submicron hot carrier reliability in CMOS ciruits.The modeling of deep submicron MOSFET’s hot carrier reliability,especially the hot carrier induced degradation in deep submicron MOSFET’s is also discussed.
摘要:The mechanism of distributed shared memory is incorporated into the parallel file system Parfsnow++,a system exhaustively using the memory of workstations.The system components of Parfsnow++ are introduced,and the file operations are presented.Finally,the performance evaluation is given.
关键词:Parallel file system;Network of workstations;Distributed shared memory
摘要:System structure of the Neighborhood Image Frame Buffer is presented,aiming at the problem of the data stream of high speed image processing,especially the storing and accessing of the neighborhood image data.Based on this,parallel storing and accessing of the neighborhood image data are realized,which greatly enhance the speed of high speed image processing.With the discussion of the character and the unique storage structure,the idea of rotating matrix and neighborhood addressing based on it is drawn forth,revealing the storing and accessing disciplinarian of neighborhood image.Specific methods of neighborhood image frame storage and concrete examples are also given in this paper.
摘要:A reliable image segmentation algorithm is proposed in this paper.Based on the assumption that the observed image is the sum of the segmentation image and the irregular corruptive noise,the segmentation image is modeled by a MRF(Markov random field)prior distribution while the observed image is modeled by a contaminated Gaussian distribution.The Bayes formulation is adopted to obtain the conditional distribution of the a posteriori distrbution of the segmentation image conditioned on the observed image,and based on MAP (maximum a posteriori) criterion the segmentation result is then obtained by applying ICM(iterated conditional mode) algorithm to maximize the a posteriori distribution.The algorithm has reliable segmentation result compared to that proposed by Lakshmanan et al .Experimental results are very satisfactory.
摘要:An efficient heuristic clustering algorithm is discussed in detail by use of deterministic annealing.The clustering problem is considered as a physical system.Finding the optimal solution to clustering problem is transformed into solving a series of problems to minimize the free energy and obtain the global minimum which varies with temperature.The examples indicate that satisfactory results are obtained for the problem to which the traditional clustering algorithms are helpless.
摘要:A novel motion estimation algorithm which is based on compensability analysis of the input video signal is proposed.The motion estimation operation is performed only on the compensable blocks in the image frame.The time of estimation computation and the bits for coding the motion vectors are all reduced significantly than those of conventional algorithms.
摘要:Microkernel technology is the newest development in operating system for the last decade.This paper discusses the architechture design of microkernel based operating system COSIX and the architechture design of system software platform COSA.It also discusses their implementation and system characteristics,etc.
关键词:System software platform;Microkernel;Server;operating system;Network software;Detabase management system
摘要:In this paper the basic principle of Digital Particle Image Velocimetry (DPIV) is presented.After the analysis of the conventional algorithm a new motion vector estimation algorithm is proposed for DPIV.The main idea is that the original images are preprocessed firstly to decorrelate,then the sequential similarity detection algorithm (SSDA) is connected with the phase estimation algorithm (PEA) to get much shorter processing time and higher precision conveniently.It makes the near real time processing possible.Although this method is proposed for DPIV,the basic idea can be used in general motion vector estimation.
摘要:This paper investigates the sensitivity distribution of two phase flow electrical capacitance tomography,and points out that the cause of low image reconstruction fidelity at the pipe center is that the former algorithm uses the same weighting factor in different flow type.It presents a new iterative algorithm using different weighting factor in different flow type.According to the result of computer simulation,this improved algorithm is able to greatly increase the imaging fidelity of any flow phase concentration.It’s proved to be better than before.
摘要:In this paper,we design a new approach of the training vector hypersphere shrinking and codebook learning formula,and present a fast fuzzy vector quantization algorithm (FFVQ) based on the learning vector quantization (LVQ) and fuzzy vector quantization (FVQ) technique.It can reduce the dependence of resulting codebook on the initial codebook selection,is not trapped in local minimum and has low operational complexity.Simulation results show that as the image coding performance of FFVQ is compared with that of FVQ,the training time is reduced more,and the PSNR value is also improved.
摘要:In this paper,we prove that when periodic signals are sampled nonuniformly,all hanmonics will not interleave with spurious spectrum caused by nonform sampling,only if the spurious spectrum does not equal to integer time of fundamental frequency of the periodic signal.
关键词:Nonuniform sample;Sampling time offset;algorithm
Sun Wenfeng, He Songhua, Guo Guirong, Li Weimin, He Zhaoqing
Issue 2, (1999)
摘要:A new algorithm for detecting target in strong ground clutter background based on one dimensional high resolution range profile(HRRP),called the Adaptive Range Cell Integration(ARCI) detection method,is presented in this paper.The ARCI shows that it’s possible to get the one dimensional range profile owing to the high range resolution of wideband signal.The space distribution feature of strong scatters can be measured with the range profile,integrating the energy in continuous range cells,thus it is effective in improving signal to clutter ratio (SCR) and then the detection performance.
关键词:CFAR;Signal to clutter ratio;High resolution range profile
摘要:This paper presents an improved algorithm for the synthesis filter bank of a MUSICAM decoder.Computational amount and memory size are reduced.Analysis of finite word length and simulation result on coefficient sensitivity are given.It also provides a foundation for the implementation of MPEG audio decoder.
摘要:A new efficient algorithm of feedforward neural networks on single parameter coordinate rotatory method is proposed,and the complexity of calculation is discussed deeply.It shows that the reason why the convergence speed of this algorithm is very quick is the complexity of calculation has close relations with the parameters of network’s nodes only.The numerical experiments prove that the algorithm is improved obviously compared with BP algorithm to solve the practical problem.
摘要:Tabu search (TS),which has been extensively applied to solving global optimization in many areas (especially in combinatorial optimization) is a very effective generalized heuristic search technique found in recent years.In this paper,a TS strategy based neural network structure,called Tabu Machine(TM),is proposed.The conclusion drawn after conducting two simulations about TM:the first is a random optimization and the second is a TSP problem,is satisfactory.The shortcoming about TM is also pointed out.
关键词:Tabu search algorithm;global optimization;Tabu Machine;TSP problem
摘要:In this paper,three procedures of optimal position location are developed in two coordinate bistatic system,which use the measurements of slant range and the azimuth referred to T/R station and the ones of range sum(target to T/R and R) and the azimuth to R station.The accuracy analysis is given respectively.The position location accuracy contours are drawn to evaluate the location performance via computer simulation.
摘要:We present a new spacial coding method named as MPG controled quadtree coding.Simulation results show that the coding efficiency of this method is a bit better than DCT,and the algorithm of this method is simple.
摘要:A low bit image coding scheme based on wavelet transform was proposed.The modulation function of the human visual system (HVS) model was incorporated into the coding scheme of vectors of high activity to enhance the visual quality of the decoded image.Experiments prove that this new scheme performs better than published schemes in the aspects of high compression ratio and high image quality.
摘要:In this paper,the mathematical model of synchronized chaotic circuit based on chua’s circuit is obtained and the average rate of convergence is calculated according to the theory of differential dynamics firstly.Then,the synchronized chaotic circuit is synthesized with OTA (Operational Transconductance Amplifier).Finally,the rates of convergence and the parameter sensitivity of synchronization are discussed by computer simulation and experiments.
摘要:The dispersion characterstics of periodic and quasi periodic multi planar dielectric gratings are analyzed by a method which combines the multimode network theory with mode matching procedure.The calculations for many grating structures show that compared with other methods presented in the literature,the present approach has the advantages of simplicity,accuracy and suitability.
摘要:A macro model of an external cavity semiconductor laser(ECLD) is presented in this paper.The model can be used as an optic electronics integrated ECLD model for computer aided design and as an analytic model for observing light feedback effects on a single longitudinal mode LD.The model permits the analysis of characteristics of an ECLD using general circuit analysis software and the analysis results are in basic agreement with reported theoretical and experimental data.
关键词:Semiconductor laser;External cavity feedback;macro model
摘要:The relationship between C V index of varactors and the parameters of materials and technics of the device has been studied on the basis of a new improved impurity profile model of hyperabrupt varactors.Then the numerical calculation and the practical relation curve have been fulfilled.
摘要:This paper presents a new approach to make application of computer software Y2K compliant.It’s called “Two digit Hybrid Radix year number for year 2000 and beyond”.In a 2 digit hybrid radix year number,the higher digit is treated as hexadecimal,displayed in a decimal like style with font patterns as 0 9 and `0 `5,while the lower digit is treated as ordinary decimal,so that the year 1900 is represented and processed as 00 while the year 2000 as ’00.The method can be embedded in an application system,so that the system can be made Y2K compliant in a systematic way.In this approach,no year number identification is needed,and the most time consuming,most costly,more unreliable portion in other approaches is avoided.
摘要:A sub region connection method is developed when we use the FDTD method in analyzing compound conducting structures.The transmission and reflection characteristics of waveguide elements and the radiation of waveguide aperture antenna are taken as numerical examples.This method has the advantages of saving computational space,reducing running time,and being flexible in setting up FDTD grid.