摘要:Router constitutes the primary factor influencing the performance of massively parallel processors(MPP). The design of a new low-cost fully adaptive minimal router (LFR) for mesh networks presented,in which a new low-cost fully adaptive minimal routing algorithm(LCFAA), an improved wormhole routing switching technoloy-elastic wormhole, and an integrated now control strategy based on virtual channel flow control are introduced.
摘要:Through the analysis of the features of the bridge object in the long-range infrared images, a new knowledge-based method for bridge recognition is proposed in this paper. We fulfill dim object detectioll and present a fast algorithm for bridge reCOgnition. The method is very effective in our experiments.
摘要:This article argues that full-parallel algorithms of traditional sigmoidal neural networks based on average fire rate which have excluded useful time operation seem improper for the invariance.Thus. we present a spiking network that intreduces a space search mechanism into neural computing. The search can transform space coordinate into time coordinate. Consequently, the operations based relative positions are converted into those based relative time which are very easy for neural realization through delay connections. Accordingly, we develop a world-centered model (WCM) that consists of a space searcher and a feature transfer vector (FTV) memory. WCM is a pure neural network that represents the neural principle of invariant recognition.
摘要:The paper introduces a classifier integrating method which transforms the problem of designing one superior classifier into designing a series of relative superior classifiers. The methed also transforms the problem of high dimensional division into different lower dimensional division. The paper proposes a Hybrid Integratd Method for character recognition in scene image, which means much noise and distortion. The method is proven to be effective in establishing practical syStems.
摘要:The research work is focused on the specific application of switched capacitor array(SCA) in high energy physics experiments. A new kind of an is proud, which can effectively overcome the influence of clock feedthrough on sample precision and offset voltage of read-out amplifier on read-out precision. An amplifier which can meet the requirement of rail-to-rail application is specially designed. On the basis of circuit simulation, full-custom layout of this mixed-signal circuit is designed, according to the deSign rule of 1. 2μm N-well double-metal double-poly CMOS technology. The test results of the sample dice show that the dice fulfills the function expected successfully, with maximum sampling rate 180MHz, co gain(sampling rate: 6MHz) 0. 99, and nonlinearity 0. 02 %.
摘要:A multi function voltage-mode sorter band on magnitude is proposed and successfully fabricated. In this sorter, switched-capacitor circuit is used to track/hold input signals. High resolution Winner-Take-All network is employed to find maximum voltage level. Then, sorted voltages based on magnitude are outputed in time-shared way. The sorter can obtain the order of input terminals on the basis of the magnitude of input voltages. In addition, it can output input voltages in a sorted order on thebasis of the magnitude of input voltages. It is fabricated in single-metal single-poly 2μm-N well standard digital CMOS process. The experiment results show that the sorter works correctly and has good perfor performance.
摘要:A new Huffman decoding algorithm is given in this paper, which is not only efficient in memory requirement, but also can speed up the process of decading. The hardware design of JPEG’sHuffman decoder is implemented as an example at last.
摘要:From analyzing the difference between MMIC and conventional microwave ciruit, theMMIC precision design method based on using commercial microwave circuit CAD software has been discussed. The emphasis is placed on modeling three kinds of MMIC elements (active devices, passive elements,and unique pattern elements determined by MMIC process) in the software. Relying on thismethod, the design and development of a Ku-band two-stage AlGaAs/InGaAs PHEMT MMIC LNA areperformed by using the conventional microwave circuit design program. The amplifier has the performance:NF of 1.66 ± 0 .04 dB with Ga of 13 .3 ±0 .05 dB at 13 .4 to 14 .0 GHZ,which is in well agreement with the CAD data and is the best result in China so far.
摘要:The femtosecond laser pulses and pump-probe method were used for measuring thetransient change of optical transmissivity of noble metal particle-matrix commpite thin films (Cu-Ba-O,An-Ba-O). It is observed that the optical absorption in the thin films rapidly increases and then recoversin a few picaseconds. optical relaxation is a process in which nonequilibrium electrons, form metallic ultrafine particles,excited by laser pulses, return to the equilibrium state. the theoretical value of electronphonon coupling constant g of An ultra fine particles in the composite thin film is got.
摘要:By using inductive metallic meshes as reflectors and couplers, submillimeter wave(SMMW) Fabry-Perot cavity lasers with a sample tube of 100cul or 20cm, 10cm in length respectivelywere designed and constructed. Then, by using NHk molecular gas as active medium and TEA CO2-10R(8) and 9R(16)line as pumping source, the corresponding SMMW lasers were obtained successfully. Finally, based on the multi-beam interference effect and density matrix theory of quantum system, the opti-cal properties and the spectral characteristics of these NH3 molecular SMMW cavity lasers were studiedtheoretically.
摘要:A matrix-geometric solution with Geom/PH/1/K queueing model for input queue analysis and a two-state Markov model for a combined output queue and HOL virtual queue analysis are Presented for evaluating a backpressure-type input/output queueing ATM switch under nonuniform traffic.Valuable results for designing an input/output queueing ATM switch using backpressure arbitration ruleare obtained.
关键词:ATM switch;Cell loss rate;Cell delay;queueing system
摘要:The caaccer freeze-out in strained p-Si1--Gex layers is studied analytically. It is found that as the Ge fraction increases, the valences effective densities of states(Nv)SiGe/(Nv)Si normlized bythat in Si decreases, furthermore, as the temperature beecomes lower, the decrease in (Nv)SiGe/(Nv)Si becomes more rapid. It is also shown that as the Ge fraction increases, although it has little effect on theionized doping concentration at room temperature, the ionias doping concentration inereases at low temperatures compared with that in Si. This implies that caaccer freeze-out is mitigated at low temperatures,which is beneficial to the operation of Sil-Gex-based devices at low temperatures.
摘要:There exists a great challenge in real-time implementation of motion estimation algorithms in vided eding because of the huge computatiollal overhead of motion estimation algorithms. A novel low-latency motion estimator architecture by using VLSI compression array is prooped for block matching algorithms in this paper, which can effectively implement full search, especially various hierarchical search algorithms, e. g.three-step search and so on. Compared with traditional accumulation treearchitecture it raises the efficiency further with more area-saving. Based on it, an adaptive block matchingalgorithm and its hardware implementation method are given, which decreases the computing burden ofblock matching to about 1/2~1/4 with a very few hardware penalty and even lower latency.
摘要:In this paper, the methods of real-time correlation operation performed by the acoustooptic Bragg cell are discussed. The mathematical models of the time-integrating acoustcroptic signal processing system that provides the correlation for spread spectrum singals with a long Cede are developed,and the design and performance of this system are described. 14. using this architecture, a desired pceossing gain of 57 dB can be achieved with a large bandwidth of 62MHz.
摘要:A novel piezoelectric plate vibration micromachined gyroscope has been proposed in this paper. Differentiating from many piezolectric gyroscopes ,it is based on a piezoelectric plate without theuse of any other elastic element. The pienzoelectric plate is excited in the polarization direction to produce thickness extension vibration mace. If there is an input rotation rate on the vibrator, an induced voltage proportional to the angular rate will be produced in the direction perpendicular to the plane composed by the polarization axis and the input angular rate axis according to the Coriolis effect. It can sense two-axis rotation rate. In this paper, the fundamental theory of this piezolectric micro gyroscope is presented. A maero stimulation experiment is described. It is shown that the experimental results are close to the theoretical predictions. borne calculating results of a given micromechanical gyroscope made of PZT film are given.
摘要:The dependence of frequency chirping on probe power is analyzed for a wavelength converter based on cross-gain modulation in semiconductor optical amplifiers. The methods to design the timun all-optical wavelength converter are proposed.
摘要:In view of the drawback of traditional bridge 4-quadrant converters which need higher dc-side operating voltage , a new type topology of converters is proposed whose dc-side voltage is reduced by approaching 50%, thus appropriatc for lower voltage applications. This paper presents the modeling with the slate space averaging method , and steady state theoretical analyses of the converter. The validity of the topology and its model along with the theory is confirmed by both MATLAB numerical simulationsand experiment results.
摘要:The challenges in further development of Si microelectronics, especially MOSFET’s into foe sub 0. 1μm regime in light of fundamental physical effects and practical consideration are discussed.The key issues of physical limitations in Si microelectronics includes: fundamental physical limitations,material limitations, technology limitalions, devices limitations, circuits and system limitations. Several potential alternative devices thai may takes us to the outermost limit-ations of Si MOSFET’s scaling down,such as single electron devices are introduced.
摘要:Sequential images play a significant role in reconstruction of 3D images for computeraided surgery and radio-therapy. The most important problem we must solve is to realize optimal and automatic boundary detection with feasible computational coat. In this paper, we develop a dynamic programing(DP)-based method with the aid of "band" and introduces a novel searching strategy to achivethis goal. Experimenal results in ultrasonic image sequences of normal and abnormal cardiac vessels prove this method to be practicable.
摘要:Protocol conformance testing and interoperability testing are important ways to encure the complete interconnection among interconnection among ATM products and services This paper analyses the test requirements of ATM, and then proposes the test architecture and test notation suitable for ATM protocols Fi. Finally, this paper presents the application of PITS in the testing of ATM adaptive layer protocol CPAAL5.
关键词:Anonymous transfer mode;Conformance tesing;Interoperability testing
摘要:A structure of integated disk array system is introduced in this paper The hardwares and the softwares of Dual Port RAM HOST interface on it are described in detail
关键词:HOST interface;Tual-Part RAM;Integrated disk array
摘要:Data organization is a key issue in the Client/Server architecutre for its impact on system performance and reliability. On the basis of experiences in client/Server application development, we discuss the logical architecture and implementation of performace-improving master/slave data replication and main-memory/disk data replication.
关键词:Client/Server;Master/slave Data Replication;Main-memory/disk Data Replication
摘要:This paper studies the transformation properties of hidden units of linear or nonlinear feedforward networks(FNN) classifiers. The mechanisms of how to organize information in the linear or nonlinear FNNs are disclosed. It is proved that the hidden output patterns in linear FNNs are correlated and the ones in nonlinear FNNs are decorrelated. Moreover, a theorem , which shows that the nonlinear networks break through the "bottleneck " in linear networks and obtain independent hidden outputs, is given. Finally,an example, i. c., parity 3 problem, is used as the simulating data and the related experimental results are presented.
摘要:The high-voltage ultra-fast photoconductive semiconductor switches(PCAS’s) fabricated with Si-GaAs coated by all-solid insulation protection technique are reported in this paper. The largesthold-off strength of 35 kV/cm of fabricated switches is obtained. We have used PCSS’s to switch voltage as high as 30 kV/cm and current about 100 A and preduced rise time as fast as 200 ps in the initiation of lock-on. Using all-solid technique to fabrication of high-voltage GaAs switch makes the device structure more simple and more practical.
关键词:photoconductive switches;Lock-on effect;Pulsed power technique
摘要:A second order large tuning-range bandpass active filter has been developed with GaAs MMIC technique using negative resistance circuits and integrated planar Schottky diodes designed by ourselves. The filter has a 200MHz 3 dB-bandwidth and a 600MHz tuning-range of centre freqency, from 1. 6GHz to 1GHz, and occupies an area of 1. 6 ×1. 8mm2 on the wafer including all biasing elements.
关键词:GaAs MMIC;Tunable active filter;Negative resistance circuit;Integrated planar schottky diode
摘要:The basic principles and ceding techniques of the waveform inteoplation for speech analysis and reconstruction are described in this paper. The practical architecture for a 2 .4kbps vocalerbased on waveform interpolation is also given. The redundancy of the pitch-cycle waveform in residual signal is removed effectively by characteristic waveforms extraction and voiced/unvoiced Separation. High quality reconstructed speech can be obtained at 2 .4kbps bit rate.
摘要:The effect of several depolarizing agents on the electrical performance of CA35 highternperature(155℃) wet tantalum capacitors was studied. The results of 1000 hours-loading test show that CuSO4·5H2O and VOSO4, as a depolarizing agent of low-voltage and high-voltage capacitors respectively, can improve greatly the electrical performance of the capacitors, and the electrical parameters of the capacitors are superior to military standard SJ20207-92.
摘要:The problem of network reliability from source s to a specified set K is considered. It presents a new topological formula which contains less terms and operations than corresponding Satya-narayanna formula based on inclusion-Exclusion Principle and Disjoint Sum Formula. Based on this formula, a new algorithm is presented which improves the corresponding Satyanaraysnna algorithm and can generate compact reliability expression.
摘要:The design keys for mm-wave low-noise PHEMT have been discussed in detail. By use of Schroedinger-/Poisson equation and the well-known device equations, numerical calculation and CAD optimization for the hetero-layer-structure parameters of the device have been completed. Therefore, theoptimized composition, doping level and thickness of the hetero-layers have been determined. Based on the analysis, the developed device has the performance: NF of 1. 92 dB with Ga of 6. sclB at 34z.4Gfu, which is the best result so far in China.
摘要:A novel image synthesis method of human face with various expressions is presented. In this method the geometrical and textural information implied in expressions are uncoupled. A 2D grid is used to extract the geometric variation of human face with different expressions. A Principal Component Analysis (PCA) is used so that the grid Sequences can be expressed by a few eigengrids. Therefore a parametric model can be developed to control the expressions effectively. During the image synthesis process, image warping and morphing techniques are employed to combine the textural information, supplied by a few reference image frames with the geometrical information, represented in the grid model, so that vivid facial images with different expressions can be preduced. The method presented in this paper has been investigated with some preliminary experiments. Its effectiveness has been proved.
摘要:CCD camera of faint light accumulation is a high sensitivity camera, which makes use of transformed accumulating control technology of CCD. The technology route of this achievement is making integral of faint light image signal, restraining random noise, utilizing the specific property of charge accumulation of CCD image seiner, changing its read-out and shift time and extending charge accumulating time of signal on the target surface, thus increasing the sensitivity of static and substatic-Objects. At the same time, a series of electric circuit and peripheral controlling devices are developed for restraining random noise, thus effectively increasing sibal-to-noise ratio of system.
摘要:Based on wavelet analysis,an approach combining iterative fractal coding and statistical prediction fractal ceding to image compression is presented. Experiment shows that bit rate=0. 236 bPP and SNR=30. 18db are achievable for the typical 512 × 512 × 8 gray image.
摘要:A novel switched-current Hamming neural network for pattern recognition is firstly proud. The calculating circuit for template matching is composed of current mirrors. Then levels of matching are compared in the switched-current sorter based on magnitude and the results are outputted.The pro-Hamming network is able to output all of levels of matching and stand patterns corresponding to these levels according to magnitude of levels. On the other hand, this network could realize the absolute rejection and relative rejection. This improves the reliability of the system. The structure of this network is simple, flexible and its scale can be easily extended. PSPICE simulation shows that the network has high resolution and high precision. Since switched-current structure is employed, this network is able to be directly fabricated in a standard digital CMOS process and easily implemented in VLSI technology.
Yeshiwei, Shi Zhongzhi(Dept of Computer Graduated School USTC Beijing 100039)
Issue 11, Pages: 140-144(1998)
摘要:Vectors, which consist of the output of every neuron in same hidden layer corresponding to different samples, should be nonlinearly correlated. With this basic fact, this Paper firstly gives the definition of linearly correlated vector and corresponding nonlinear correlation measure for every hidden layer, then adds or deletes a neuron for a hidden layer according to its nonlinear correlation measure and adjusts the neural networks weight, values appropriately. This method can not only avoid confining the number of neuron units in a hidden layer, but also escape local minimum during the learnig process. According to error analysis in detail, if gives the optimistic rule of deleting neuron in hidden layer. Numerical experiments illustrate its efficiency.