In this paper, a novel idea for optimization of image segmentation algorithms based on algorithm evaluation is proposed. According to this idea, a segmentation expert system is designed and realized. Experimental results show that such an idea is feasible and such a system is workable. This also provides an effective way for the optimization of various image processing and analysis algorithms.

Sub-image information of the wavelet decomposed finger print image is essential to the quality of reconstructed image. This paper describes the importance of the sub-images according to the features of ringer print images, and provides a new method that can be used to compress texture images. It can simplify and accelerate the algorithm with a better reconstructed image quality as well as a larger compress ratio.

A new predictive search (PSA) algorithm is proposed in this paper, where the search displacements are between -16 to +15.5. It firstly derives motion vector from the linear weighting of motion vectors of the three adjacent blocks, then searches from the prediction point using 3× 3 search window and search with the step of one until it reaches the boundary of search area or the local minimum point lies in the center of search window. Based on the spatial characteristics of center-biased distance between motion vector and corresponding prediction vector and temporal correlation of motion vectors, this method can significantly reduces the search times. Simulation results show that the proposed PSA is efficient to reduce the search range, search times, and computational complexity. Comparisons of PSA with other fast search algorithms are also given in detail.

The concept of instantaneous frequency(IF) and its traditional construct method--analytic method are introduced in this paper.Then by means of joint time-frequency analysis,the fphysical meaning of IF and its relation with the time-frequency distributions are also discussed.It is pointed out that one of the destinations of 2-D time-frequency distributions is to estimate the IF,and the different distribution can be of different effect.The performances of several popular distributions are compared,consequently AOP method is found to be a suitable one to estimate the IF.

Wavelet neural networks (WNN) is a powerful tool for function approximation. In this paper a new model named adaptive time delay WNN(ATDWNN) is proposed which combines time delay neural network and wavelet decomposition. ATDWNN is used to approximate signals having different time delays in the same class. In order to train ATDWNN, time mechanism based competition learning is also proposed. It is shown through experiments that ATDWNN can not only approximate signals having different time delays by the same superwavelet, but also detect these time delays successfully.

In this paper, the estimation of attraction region for a given equilibrium position of associative memories is given with Lyapunov function method. It is very important for the implementation of associative memories and optimization with neural networks.

The performance of DOA estimation will be distorted in the presence of dielectric dome in the Dielectric Dome Array(DDA) because the dielectric dome diverge the planar wavefront into non-planar wavefront inside it. We proposed a far field approximation(FFA) method in this paper to improve the performance based on DDA. This method first reconstructs a FFA covariance matrix, which is Toplitz and approximate to the far field data covariance matrix, and then the high-resolution eigenstructure techniques can be utilized to estimate the DOAs without revising the planar wavefont assumption. Simulation results confirm the theoretical work.

A kind of Chinese combined language model,that takes into account POS(part of speech)information in a trigram-based statistical language model, is presented in this paper. The theoretical analysis and experiments all show that the model not only is lower than trigram model in PP(perplexity), but also is superior to trigram model in dependence on test text domain.

The relation between weight and sidelobes of cross-correlation function and the relations between Hamming distance and numbers in a sequence are systematically studied in this paper. And then the general-purpose weight theorem and Hamming distance theorems are proved. Furthernmore, the transformation relations of weights, Hamming distances between the orthogonal codes,complementary codes, and the code pair of standard orthogonal transformation in the procedure of Tseng dissolving, and the association between them and runs are studied. And then, a series of importance corollarys are obtained.

This paper presents one certain chaotic radar waveforms based on one-dimensional chaotic sequences.This paper derives theoretically the formulae of the ambiguity functions of two casess,i.e.,(1) continuous chaotic radar waveforms;(2)rectangular pulsse raaar waveforms modulated by cut-off chaotic sequences.Such theoretical analysis and numerical simulations denote that the ambiguity functions of the first case are shaped by iedal thumbtacks,and those of the second case are shaped by sharp knife edges.These results hint that such chaotic radar waveforms have good range and velocity resolutions.

In this paper, the main schemes of connection admission control (CAC) in ATM networks are briefly overviewed especially the principle of dynamic bandwidth allocation. Then the fair share of the bandwidth among different traffic sources is analyzed based on cooperative game model. A CAC scheme combining genetic algorithm (GA) to optimize the bandwidth-delay-product formed utilization function that ensures the fair share of accepting/rejecting the incoming calls is proposed. Simulation results show that our scheme ensures fairness of the shared bandwidth to different traffic sources.

Firstly, some reduction operations of marked T-graph with weights are given in this paper. Then the conditions for preserving liveness of marked T-graph with weights after these operations are discussed. These conclusions have important significance in analysis and synthesis of marked T-graph with weights.

The relativistic effects on point-positioning precision in the GPS are studied with reference to an earth centered inertial coordinate system. First, the metric coefficient 900 in the earth centered inertial system is obtained from the sun s Schwarzschild isotropic metric, with the gravitation of the earth and the moon taken into account. Then the general and special relativistic effects on the GPS atomic frequency standards, the GPS satellites orbits and the propagation of the GPS navigation signals are estimated, based on which the point-positioning errors caused by the gravitation of the earth, the moon and the sun are evaluated, and the sources and importance of each components of the errors are analysed.

In this paper, a novel computer-aided autodevelopment method of analog integrated circuit macromodel is presented. The new method can be employed generally, and the macromodel autodeveloped is accurate. The autodevelopment procedure of macromodel is described in details, and the relevant software flowchart is given. Then some researches and new ideas in the critical algorithm are presented. Finally, a design example using our new approach is presented.

A noise network model of single-mode semiconductor laser diode is established based on the Harder LD small signal noise model and straightforward calculation of rate equation in the TMS software. The characteristics of frequency response and noise are analyzed by using the presented model.

Making use of the linear field theory, the dielectric Cherenkov maser with an annular plasma is analysed in detail. The beam-wave interactions corresponding to, (a) the thin annular relativistic electron beam (TAREB) surrounding the annular plasma, (b) TAREB within the annular plasma, and (c) TAREB inside the annular plasma, are discussed; and the dispersion equations are derived, respectively. Numerical solutions for the dispersion equations with a complex angular frequency are directly carried out. And the cutoff frequency, the operation frequency and the wave growth rate in the slow-wave system are obtained. Finally, the effects of related parameters on them are presented.

Dielectric conformal dome phased array antenna uses the dielectric dome to change the direction of the scanning beam. It is supposed to widen the scan-range of planar phased array to half spherical space or more. It is suitable for missiles and air-borne radars. When analysise its scanning characteristics, the divergence factor(DF) of the dome must be calculated. This paper provides two methods to study the DF which are based on geometrical optics, and compares them. These methods are very useful in studying the reflection and the refraction of the non-planar layered dielectric.

In this paper the theoretical basis of cellular neural networks first created by L. O. Chua(1988) are discussed. Some fault are shown, new research method are suggested.

In this paper, fractal image coding and fractal interpolation for image compression are studied at first. Because the reconstructed image quality using the two methods at high compression is not good, a new image compression idea combining these two methods is proposed. The coding results verified the new idea. The compression ratio is as high as 76 to 1 as PSNR being 28.5 dB and coding time is very short.

In this paper, lapped orthogonal transform(LOT) is applied to orthogonal frequency division multiplexing(OFDM) system for modulation and demodulation, and a new OFDM scheme is proposed. The performance comparison between the new scheme and traditional OFDM system using DFT is given. It shows the new scheme can decrease bit error rate(BER) efficiently in wireless communication.

The spatial Fourier decomposition method is a new method for the SAR imaging. It incorporates the spherical nature of the radar s radiation pattern and so the image would not distort even at near field. This paper presents an algorithm for the spatial Fourier decomposition method. The imaging results of simulated targets prove it is effective.

A method for rapidly determining long generation lifetime under linear voltage sweep was presented. The C-t transient curve needs not to reach saturation. The data processing is simple. Furthermore, it needs not to know the sample s impurity density.

Two universal integrator-based CC (Current Conveyor) network configurations are presented. The architectures proposed are constructed using a canonical FLF (Follow-the-Leader Feedback) configuration with an input distribution and output summation CC network. By setting any particular input distributing and output summing passive element parameters,the arbitrary high- order transfer functions can be realized.General explicit design formulas are derived. It is also straightforward to obtain the universal filter architectures and the corresponding design formulas of any particular order.