A new time-frequency representation called Dopplerlet transform, which uses the dilated, translated and moduiated windowed Doppler signals as its basis functions, is introduced. It is proved that the matchihg pursuits based on Dopplerlet basis functions are convergent, and that the energy of residual signals yielded in the decomposition process decays exponentially. Simulation results show that the matching pursuits with Dopplerlet basis functions can char-acterize compactly a nonstationary signal.

In this paper, the recovery of transient signals embedded in noise is studied. It is noticed that there is different feature between the transient signal s Gabor coefficients and the noise s Gabor coefficients distributed in time-frequency plane, the Gabor series method is presented for recovering transient signals embedded in noise in terms of hypothesis test theory. The computer simulation is implemented for exponential transient signal. This method is also used to process real-life signals.

This paper presents a ESPRIT-based algorithm to estimate carrier frequencies of independent spatial signals with no geometric constraints on the array. After mapping the element space data into DFT space, the "frequency vector" is extracted which is independent of array information, and then beamspace ESPRIT is used to estimate the spatial signals frequencies via an indirect ESPRIT method. The simulation results demonstrate our algorithm to be of high-resolution and high-precision with arbitrary array geometry.

Based on the Yule-Walker equations, a traditional Maximum Entropy Method (MEM) is the Levinson-Durbin recursion algorithm. The algorithm omits errors that are pro-duced when estimated values of autocorrelation sequence take the place of their ideai values. A new algorithm is presented in order to decrease the errors as much as possible. The algorithm improves its performance of spectral estimation and reduces computational complexity by means of the clues of some existing algorithms. The results of simulati...

This paper presents an algorithm for coding color video signal using 3D-wavelet transformation. When the 3 color channels of RGB of the video signal are converted to color channels of YCbCr, and" the frame order t of the signal is replaced by order z, this signal can be considered as 3 blocks in a 3D-space. Each block presents a color channel. After splitting each block into smaller sub-blocks, imitate the method of 2D-wavelet transformation for image, the sub-blocks can be transformed with 3D-wavelet. Most of energy of each color channel is in the transformed low-frequency bands. These bands affect the visual quality of the video signal most. After quantizing different bands with different precision and entropy coding each band, the intra- and inter-frame redundancy of the color video signal can be eliminated and the data can be compressed. Our simulation experiments show that this algorithm can obtain very good results.

An unsupervised clustering algorithm is described in this paper, which utilizes the coefncient of decomposition of the Mueller matrix of the polarimetric SAR image. The algorithm can classify the image into three distinct categories, i.e., building area, vegetated area, and slightly rough surface (e.g. water) without any terrain measurement according to the various experienced knowledge about scattering mechnism of terrain targets. Compared with other unsupervised clustering algorithm based on the single polarimetric gray-scale image, this algorithm is characterized that it can not only cluster every pixel according to its character, but also determine the scattering mechnism of every class, and the type of targets.

Prefiltering is indispensable for multiwavelet transforms. This paper proposes a prefilter with the length of l and an approximate order of 2 for GHM multi-wavelet, and proves that the order of prefilters for length 1 repeating signals cannot be or greater than 2 for GHM multi-wavelet, after analyzing the constructing ideas of various currently existing prefilters for GHM multi-wavelet.

This Paper is focused on the research of connecting the HLS(high level synthesis) result to the technology of IC. A knowledge-based high level technology mapping methodology is proposed, based on the traditional technology mapping theory. The method is discussed from the views of knowledge representation, knowledge acquirement and knowledge utilization, including: (1) initiating an expanded production representation approach for the knowledge of circuit structure; (2) presenting a technology mapping knowledge acquirement technique based on VHDL; (3) giving the controlling tactics and algorithms in knowledge utilization; (4) presenting a half-automatic method for KB(knowledge base) maintenance, with which, the contradictory and the redundancy in KB can be found; (5) raising a practical method that embeds the algorithms into the knowledge-based system, in order to reduce the complexity of the KB. This system has been connected to 3 kinds of IC production line, and this verifies the theory and method of the paper.

This paper proposes an efficient on-line learning method for radial basis function (RBF) neural networks. The proposed learning method not only dynamically allocate the network resource in accordance with the increase of input Information, but also efficiently recycle the redundant resource of the network. During the learning process the parameters of the network can be sequentially adapted. The learning criterion, mechanism of increasing and decreasing resources and the parameter adjustment algorithm are elaborated. Meanwhile both the mapping approximation ability and predication performance of the network are analyzed in details.

In this paper, the effect of input noise on the typical stochastic Hopfield neural network modei is discussed. It is shown that the expectation of the stochastic HNN of the trajectory is uniformly bounded over time. For practical design purposes, the stochastic input error estimates for the stochastic HNN with respect to the corresponding deterministic HNN is derived. In addition, the designer can use these results to constrain the design space so that the achieved design satisnes the performance specifications whenever possible.

The operating principle of using scaning force microscopy(SFM) to measure surface electrostatic potential is described in this article, and the block diagrams of different mode and a few experimental results are given simultanously.

Based on the working characteristic of Schmitt circuits which have two detection thresholds to input signals, this paper proposes that Schmitt circuits have the same sequen-tial characteristic as flip-flops which are used as memory units. With the design procedure of sequential circuits, varieties of Schmitt circuits based on integrated gate circuits are systemati-cally investigated, and some new designs are found. PSPICE simulations prove that the circuits designed in this paper have ideal characteristic of Schmitt circuits.

Known Bits(KB) method is a way of increasing the performance of turbo code by inserting known bits periodically in a code word without increasing in complexity. KB method normally causes uneven protection due to the process of random interleaving. This problem can be solved with a KB interleaver to ensure uniform distribution of KB. It is found that KB interleaver can improve the performance of turbo code with pseudo random interleaver or S-random interleavers. This method also makes a rate-matching to be easy which is fairly important for systems using turbo codes.

An ML Classification algorithm that classifies the terrain object in the multi-band, full-polarization SAR image is described in this paper. Its main feature is that the optimal state of polarization SAR image is utilized to classify objects. The searching algorithm for the optimal state and the classification algorithm of terrain targets are provided, and the classifier s performance is verified using the multi-band (P, L and C band), full-polarization testing data that is acquired by AIRSAR. Compared with the single band, single polarization SAR data, the classification accuracy of the optimal state based classification algorithm is improved significantly.

An algorithm of TDOA location using plane intersection is introduced. Its advan-tages are closed-form, no need of iteration and high accuracy. A modified algorithm is given also to save computation with simulation results.

This paper presents a new method of designing high-order CCII filter.The method is synthesizing input immittance of passive ladder network by CCII gyrators and realizing high-order current mode CCII nlter.The paper analyzes non-ideal character of CCII and gives out the compensation method. Compared with other similar circuits, the designed circuit using this method has simpler structure and fewer number of CCIIs. Design example of 6-th order Butterworth filter and its computer PSPICE imitation results before compensation and after compensation are also given.

Some hot-spots of research in the field of the field emission array (FEA) and the vacuum microelectronics (VME) have been developed abroad in the last phase of 1990s. The expand speed of the FEA subject is very fast, and the non-heated electron emission sources iriside some advanced electron-physics equipments have also got a relevant development during this period. Considering those trends in the fields of basie science-FEA and VME abroad, according to the particularity of the vacuum electron emission, to seize the centralized relatively cathodes in the research or the report, a brief summary, review or comparison has been made, a more important type of the field emission array-the plasma electron source array which has already been developed into practical use recently in the information display has also been introduced simply.