摘要:Satellite ATM(Asychronous transfer mode) is a hot research area of satellite communication in recent years and congestion control is one of the key issues.This paper focuses on the ABR(Available bit rate) congestion control algorithm in the satellite ATM network based on onboard processing,which is implemented by using the traffic prediction.Owing to the long propagation delay of satellite channel,the accuracy of prediction decreases and it degrades the performance of ABR congestion control.In order to overcome the impact of inaccuracy,a new method 'Long-time prediction plus dynamic control bandwidth utilization factor’ is suggested.Simulation results show that this method has advantages on the cell loss probability,link usage,fairness,adaptation to different number of TCP(Transport Control Protocol) sources and transient response.
摘要:The paper investigates a new blind multiuser detector called constant modulus detector and analyzes the capability of the detector to suppress multi-access interference (MAI).The results show that the performance of detector is poor when the received amplitude of the demodulated signal is less than the critical value 1/ 3 ,contrariwise,the detector possesses stronger capability of MAI suppression.The simulation examples demonstrate the results.
摘要:A real time pitch detection scheme is proposed in this paper in order to reduce the delay of the existent pitch detection algorithms. The experimental results show that the new scheme can obtain exact pitch level within the current frame and reduce the delay of the existent auto-correlation pitch detection algorithms using smooth processing.
摘要:This paper proposes a novel method for vector coding,which approximately represents a vector with its orthogonal projection on a straight line.It encodes every vector with 3 codes,where two are used to represent two reference points to determine a straight line,and the third one is used to describe the location of the projection point.Since a N-sized vector quantization (VQ) codebook can determine N(N-1)/2 different straight lines,the new method can encode vectors with very high coding accuracy based on a rather small codebook.Theoretical analysis and experimental results suggest that its coding accuracy based on a N-sized code-book of size N N2 is comparable with that of VQ coding based on a N2-sized codebook,and much better than multi-stage VQ based on two codebooks of size N,while its computational complexity is much lower than the latter ones under equivalent accuracy obtained.The method would be a powerful tool for vector coding with high accuracy requirement.
摘要:In this paper,a novel method to design the Daubechies filter is introduced.Based on this,an algorithm of the image perfect decomposition and reconstruction is proposed.In fact,for orthogonal wavelet transform,the padding approach to the boundary condition is periodic.The approach is derived from the design method of Daubechies filter.When the signal length is more than the length of the filter,the signal can be decomposed and reconstructed perfectly.
YANG Shao-qing, ZHANG Xin-hua, XIAO Ming-jie, ZHAO Chang-an
Vol. 28, Issue 10, Pages: 20-22(2000)
摘要:Fractal dimension is an important parameter of a chaotic signal.It is usually to determine a linear scaling region when evaluating fractal dimension,however,determining the linear scaling region is still a challenging problem.The accuracy of the fractal dimension estimated will be directly affected by the accuracy of the linear scaling region.This problem is deeply developed here and solved better.Based on the GP method,a series of algorithm for computing the correlation dimension D2 is proposed.This new algorithm needs only a little computational time,has high accuracy and stronger ability of resisting noise and it can automatically calculate the correlation dimension.
关键词:chaos;reconstruction;the largest Lyapunov exponent;linear scaling region;correlation dimension
摘要:In this paper,we analyze the principles of deterministic annealing technique and competitive learning algorithm for image coding,and present a new stochastic competitive learning vector quantization algorithm for image coding.It combines the procedure of competition with the minimization of cost function.Simulated annealing is used in learning procedure,and several new strategies are presented in the vector quantization for image coding.The algorithm can eliminate the effect of initial codebook selection on the quality of clustering,is not trapped in local minimum,has a good convergence rate,and can get the codebook with good performance.Computer simulation results confirm the effectiveness and robustness of the approach.
LI Yun-song, WU Cheng-ke, ZHANG Zheng-yang, DUAN Yong
Vol. 28, Issue 10, Pages: 27-30(2000)
摘要:A new remote sensing image coding scheme based on the biorthogonal wavelet transform and revised SPIHT algorithm is proposed according to the statistical characteristic of remote sensing image data.The experimental results show that the proposed scheme could largely speed up the encoding process and reduce computational complexity on condition that reconstructed image quality is sacrificed little.Meanwhile,this scheme has advantage over other algorithms in real time encoding.
摘要:A new filter—Fuzzy Detection Weighted Mean filter is presented in this paper.Based on the histogram of corrupted image and its statistic characteristic,we construct the fuzzy membership functions and present the design methodology of FDWM combining new detection algorithm.For salt-pepper impulse noise image,when the noise probability exceeds 30%,FDWM gives superior performance compared with conventional filters,especially when noise probability is more than 50%,the advantage of FDWM is more prominent than other filters.This is showed by the subjective vision and peak signal-to-noise-rate (PSNR) or mean square error (MSE).
摘要:Point-pattern matching is an important problem in the fields of computer vision and pattern recognition.In this paper, a new algorithm is proposed to solve the problem of matching two point sets with the same cardinality under a perspective transformation.This new algorithm is based on epipolar geometry and genetic algorithm.Firstly an approximating solution is computed using the algorithm for solving point-pattern matching under affine transformation.Then considering this approximating solution as the initial population,a genetic algorithm is used to improve the approximating solution.Theoretical analysis and simulation results show that the new algorithm is effective.This new algorithm is also a new attempt to solve the 3-D point-pattern matching problem under perspective transformation.
摘要:The research of intelligent machines has been an increasingly concerned subject these days.Due to the great complexity of the subject,however,the research faces many difficulties for the time being.From a long period of study we discover that the foundation and the nucleus of intelligence roots in the consciousness and the mechanism of consciousness emergence is just the same as the process of comperehensive information production.As a result of this discovery,a theory and the related functional model of consciousness based on comprehensive information theory are established,and the technical feasibility of the model implementation is also discussed.The results reported in the paper may lay certain foundation for the exploration of the mysterious consciousness and a new ,and a promising,approach to the research of intelligent machines may also hopefully be initiated.
关键词:intelligent machines;comprehensive information theory;consciousness machines;inductive learning;neural networks
摘要:In this paper,Sluzek's method is investigated.Its advantage lies in decreasing moments' variance and in making full use of discriminating difference between two classes as possible.Its disadvantage is its inability of discriminating two similar classes whose differences lie in the middle.In this paper we will prove in a general form that there is much chance of improving this method,and present a new method.The prototype object is partitioned into n rings of equal moment,and the moment invariants corresponding to these circles are computed,from which some are selected as the descriptors of the objects.Finally,the improving effect of our method is verified by some examples.
关键词:moment invariant;Sluzek method;partitioning of equal moment;pattern recognition
摘要:Generalized algorithms are proposed for linear-phase paraunitary filter banks with pairwise-mirror-image frequency responses.The proposed algorithms promote fast convergence of design process.When the number of channels is even,a generalized linear factor is proposed that has additional free parameters and helps to improve the characteristics of the filters.When the number of channels is odd,a new structure is proposed.
摘要:By studying the discrete Fourier transform properties of the band-limited digital signal,the authors introduce alternating projection neural networks into the paper,expand APNN's application scope from real field to complex field,and present several important conclusions on APNN.Analyzing and discussing network's tolerance to noise,convergence rate and the spectral leakage problem of the truncated signal expected to be extrapolated by using these conclusions,the paper presents an extrapolation algorithm for band-limited signals based on alternating projection neural networks.A lot of simulation experiments show that the algorithm is effective.In addition,the algorithm is also effective to spectrum extrapolation.Owing to adopting network structure,the algorithm is prone to parallel computation and VLSI design,and consequently can satisfy real time military processing needs.
摘要:Starting with the construction conditions of basic wavelet (or mother wavelet),three topics are studied in this paper.Firstly,the numerical computation stability of dyadic wavelets is discussed with a new concept——the stableness,which enables us to measure the stabilizing ability in numerical inverse of wavelet;And then the stableness of a novel class of complex analytic wavelet——Y-wavelet (whose quality factor is changeable)is analyzed.Finally,we investigate the orthogonal condition of Y-wavelet.Theoretical analysis and results of numerical simulation show that:The stableness and frequency selectivity of wavelets contradict each other in wavelet analysis,but this problem can be mitigated to some extent by changing the scaling dilation parameter;The stability region of dyadic Y-wavelet class is very narrow and limited by larger quality factor;It is impossible to construct an integral translation orthogonal basis by using Y-wavelets,except that the real Y-wavelet might exist on some integral translation orthogonal basis.
摘要:The concept and the model of the multiwavelet neural network are proposed.Its universal and L2 approximation properties as well as its consistency are proved,and the convergence rates associated with these properties are given.Theoretical analyses and the experimental results show that multiwavelet networks can have obvious advantages in approximation performance over simple wavelet networks,and similar works have not been seen at home and abroad.
摘要:In this paper we present a cyclotomic approach to the construction of all binary duadic codes of pm lengths,by which all binary duadic codes of pm lengths are enumerated.
摘要:In this paper we investigate the crosscorrelation function Cd(t) of m sequence on finite field GF(p) and its decimation sequence,and get the following results:(1)For decimation factor d= pn+1 p+1 + pn-1 2 ,p≡3(mod 4),n odd,we show that |1+Cd(t)| 1+p 2 pn ,and therefore solve an open problem raised by Muller in [1] and generalize the result proved in [1];(2)For decimation factor d= pn+1 p+1 ,n odd,p≡3(mod 4),we show Cd(t)∈{-1,-1+ pn+1 ,-1- pn+1 };(3)In the two cases above,we study the distribution of |1+Cd(t)| with respect to t,and obtain that when p is big enough,the probability of |1+Cd(t)| achieving the maximum is very small.
摘要:A novel fast soft-decision decoding algorithm, called variable-threshold Chase algorithm (VTC),is described.The set of test patterns is first produced by using A* algorithm,then is partitioned into the equivalence classes.During the partition,the equivalence class representatives can be generated.A variable threshold is derived for optimality test on a candidate codeword.With set partition and optimality test,a fast soft-decision decoding algorithm is obtained.Simulation results show the advantage of the proposed algorithm over soft-decision decoding algorithms available is much faster decoding speed without loss of decoding performance.
摘要:This paper presents a methodology to evaluate the signaling system loading for personal communications that are independent of assumptions.We use this methodology to evaluate the signaling system loading of GSM and IS-41 system.
摘要:This paper presents an admission control scheme for providing Quality of Service (QoS) including delay tail distribution and loss probability in high-speed networks.The relationship between resource allocation for flows modeled by on/off Markovian fluids and statistical QoS is established based on an analysis of the Markov process.Using this theoretical result,the feasible test method of admission control for QoS provisioning is put forward.Our proposed admission control scheme is optimal in that it takes account of the relative precedence of flows and represents a requirement for maximizing network revenue,while providing delay tail distribution and loss probability guarantee for multiple classes flows.Its performance is investigated through numerical results.
摘要:The performance of Spread ALOHA Multiple Access (SAMA) is investigated,which combines virtues of pure ALOHA and spread spectrum technique.A more general expression to calculate throughput performance of ALOHA system is obtained.The result of research shows that this multiple access method is one of the promising candidates for the digital wireless network in the future which has the coming demand for supporting various traffics and QoS.
摘要:A new multiple access protocol which can be used in initialization process for S-CDMA system based on HFC CATV networks is proposed by comparing of some existed protocols.As in other CDMA systems,the power control also makes a great effect to the capacity of HFC networks.A complete initialization scheme is given,in which the multiple access protocol and power control are considered together.
摘要:Three connection management approaches in the multi-vender network environment are analyzed and compared:the TMN (telecommunication management network )-based approach,the TINA(Telecommunication information network architecture)-based approaches,and mobile agents based approaches.
摘要:A major cause of errors in automatic speech recognition (ASR) systems is the inaccurate detection of the beginning and ending boundaries of test and reference patterns.Accurate determination of endpoints of speech is not very difficult if the SNR is high.Unfortunately,most practical ASR systems must work with a small SNR,and the conventional speech detection methods based on some simple features such as energy cannot work well in noisy environments.In this paper,cepstrum is used as the feature to detect the voice activity.Two algorithms for endpoint detection of noisy speech signal are proposed.The first one takes the cepstral distance as the decision thresholds instead of short-time energy.The second approach modified the HMM-based speech detector to make it adaptive to the change of noise.The experiments show high accurate rates can be obtained.
摘要:In the training phase of HMM system,due to the cumulative production,the evaluation of the forward and backward probabilities in Baum-Welch algorithm needs a large dynamic range which will exceed the precision range of essentially any machine.This can be resolved by multiplying the forward and backward probability with scaling coefficients in single observation case.In multiple observation case,generally the output probability terms are introduced for each observation,which will cause the overflow problem again.In this paper,the cause of overflow problem is studied and a revised BW algorithm specially for multiple observation is deduced by redefining the optimization object function from cumulative production to sum of logarithms.The revised HMM parameter re-estimating algorithm is more stable and effective,and the overflow problem is eliminated.
摘要:This paper firstly describes a novel perceptual wavelet and its transform of speech,which is based on Bark frequency scale concept coming from the speech perceptual experiment.In the nature of physics,it consists with human's cochlea filter properties;in the sense of mathematics,Bark wavelet transform is a kind of non-orthonormal but overcomplete,reversible and self-inverting wavelet transform.When it is applied in speech recognition,small analysis frame-length is used by using the property of local base of wavelet transform,which can improve the time-domain resolution of front-end process.Experiment shows that the formed novel feature is superior to MFCC in the capability of distinguishing the consonants.
摘要:Video object segmentation is an important task in object-based video coding.In this paper,an automatic video segmentation scheme based on higher-order statistics (HOS) of displaced frame differences (DFD) and mathematical morphological operators is studied.In this approach,a map (mark) overlapping object is first gotten by thresholding a group of higher-order moment maps of DFD,and then a morphological erosion from the boundary of the mark up to the boundary of the object is performed to get the final segmentation map.A simple but efficient algorithm for separating head and shoulder is proposed based on searching for the greatest degree turning-point in head-shoulder contour curve.Based on segmentation,the MPEG-4 based video coding system we implemented with software is described.Then,a very low bit-rate video coder with object-oriented bandwidth allocation (OOBA) technology is proposed.Compared with conventional frame-based video coder,the OOBA-coder can improve subjective quality of images at low bit-rate,while decrease PSNR just a little.
关键词:video object segmentation;head-shoulder separation;MPEG-4;object-based video coding;low bit-rate video coding
摘要:Watermarking is a technique for labeling digital pictures by hiding secret information into images.Watermark embedding is a method to discourage unauthorized copying or arrest the origin of the images. In this paper, a new method is proposed.The watermark is processed as a visually recognizable pattern—binary image,which includes more information than the traditional bit streams and is "extracted" instead of only "detected" to characterize the owner.The watermark is hid in the host image by selectively modifying the middle-frequency part of the host image in conjunction with the human visual model and the image cosine transform.Experiments show that this method can survive image cropping and image compression, and get better results, and this is also a prospective method.
关键词:digital watermark;HVS model;DCT transform;radial frequency
摘要:The paper discusses shortcomings of adjacent frame difference.Based on the image energy,high order statistic(HOS),prediction theory and block processing techniques,background of high-speed dense visual target motion sequences is reconstructed,and motion states are studied.Our DSP based platform tests indicate the provided algorithm performs excellently in real visual sequences motion analyses.
关键词:high-speed motion;background reconstruction;dense visual target scene;recognition and tracking
摘要:In this paper a new application of genetic algorithm (GA) in FIR filter design with frequency sampling method is presented.As the value of transition bands sample is obtained by GA,the problem that the value obtained by searching for table method is not the optimal can be solved.Suitable improvements are made for the standard GA in the paper.Two examples of lowpass and bandpass FIR filter design are provided.Experimental results have shown that the performance of the filter,which is designed via GA,is improved as compared with searching for table method.
YANG Xue-zhi, YAN Pu-qiang, YAN Lei, WANG Jian-min
Vol. 28, Issue 10, Pages: 121-123(2000)
摘要:In this paper,the least square deconvolution algorithm is deduced from the orthogonal projection method in Hilbert space.The solution is the optimum one in the sense of least square.Using this method,the dynamic deflection signal is successfully obtained from the signal of inertial transducer,which is distorted seriously.The comparison with other methods proves the effectiveness of this method.This method provides a new approach for the measurement of dynamic deflection of bridge.
关键词:least square deconvolution algorithm;orthonormal projection;measurement of dynamic deflection
摘要:This paper deals with the closed-form recursive estimation for parameters of almost periodic moving average (APMA) signal using higher order cyclic-statistics.Two groups of representation formulas relating the coefficients of APMA signal and odd order time-varying cumulants are established,from which two direct algorithms are presented for estimating parameters of APMA signal.As a consequence,the closed-form recursive formulas of APMA coefficients are derived.Simulations are given to illustrate the performance of the presented methods.
关键词:almost periodic MA signal;time-varying cumulant;parameter estimation;closed-form recursive estimation
摘要:This paper introduces both the method of the spectrum-coding transmission of digital signal and that of wavelet threshold de-noising in detail,and also studies the performance of the method of wavelet threshold de-noising used in the bipolar linear-phase coding.To find the best scheme of wavelet de-noising,various parameters that influence the performance of wavelet de-noising are studied in this paper.In the end,the statistics of the operational volume of wavelet de-noising is made.The result indicates that the wavelet transform has practical value in the performance of de-noising and real-time processing.
关键词:spectrum-coding;method of threshold;the bipolar linear-phase coding
WANG Jin-gen, GONG Shen-guang, WU Zheng-guo, TANG Jin-fei
Vol. 28, Issue 10, Pages: 130-132(2000)
摘要:Based on the principle of alternative projection,the paper presents an algorithm of signal reconstruction from modulus maximum of signal's wavelet transform based on radial-basis-function (RBF) network.The algorithm accelerates convergence and improves reconstruction precision by taking advantage of RBF network's good linear property which occurs while the width of the basis function is very large.The algorithm can not only reconstruct signal's edges very well,but also recover signal's climax precisely.A lot of simulation experiments show that the algorithm is also effective to different kinds of signals and has characteristics of both fast convergence and high precision.
摘要:A new method for training recurrent neural network (RNN) has been proposed.By introducing the hidden representation or hidden variables into RNN,training the complicated RNN is decomposed into training a set of single neurons and a linear output layer.Based on linear approximation of RNN hidden units,RNN is remodeled with a "mixture of experts"(ME) model.Morever,training RNN is also changed into a maximum likelihood estimation of the linear systems with hidden variables.Finally,training RNN is fulfilled with the expectation-maximization (EM) algorithm.
摘要:Wavelet denoising of Donoho and Johnstone in multiresolution approximation has been evolved to become very powerful tools for digital signal processing.But the determining of the parameter jo,the low-resolution cutoff,is difficult in general case.It needs some regularities of the source signal.Which jo is the best for a given unknown signal and a noise power σ?In this paper,we construct a statistic for looking for the best jo adaptively.We analyze the properties of the statistic and give many simulation results.This adaptive method makes the scope of application of the wavelet denoising more widely.
摘要:With the growing popularity of electronic commerce,electronic lottery will be one of the necessary businesses in the Internet.One "on-line" lottery scheme based on bit commitment is presented,which has the advantage of fairness and that it overcome the limitation on time.
摘要:Hwang and Ku introduced a concept of reparability feature into key distribution protocol.In this paper,the key distribution protocol,proposed by Hwang and Ku,is analyzed.This paper also presentes security problems in Hwang and Ku's protocol,and furthermore,the argument on whether the protocol has the perfect reparability feature.So,an improved reparable key distribution protocol is proposed in this paper,which solves these security problems.A method,hiding the duplicated event markers,is also proposed and is applied in the protocol to successfully solve the perfect reparability problem.