In this paper, we propose a processor allocation algorithm using the PGG(Packed Gray code Group) for the MIMD hypercube. The number of k-D subcubes in an n-cube is C(n.k) en-k. When the PGG is employed in the processor allocation, C(n, k) PGG's are required to recognize all the k-D subcubes in an n-cube. from the simulation we find that the capability of processor allocation using only 40% of C(n, k) PGG's is about the same as that of the allocation using all the PGG's.

In this paper, we propose a new method to desing IIR Nyquist filters. By thismethod, we can design IIR nyquist filters which prossess linear phase characteristics andare lower in order than FIR Nyquist filters. WE derive expressions for zero-phased IIR Nyquist filters and propose new desing methods for them. An iteration process is used to design the proposed IIR Nyquist filter, and in each iteration step a modified version of the Remez exchange algorithm is used. Finally, the propose desing method is demonstrated through design examples.

Lossless system has been employed to design the perfect reconstruction filter banks and has particularly a close relation with the desing of orthogonal wavelet filter (OWF). With such a relation, we generalize 2-band OWF to the M-band OWF which has an improved performances. The improvement is achieved using the two techniques. One is that the wavelet low pass filter has (N-1)th order regularity with extra zeros while the existing filter has N-th order regularity. The other is that unitary matrix is designed by adding the zeros to the proper positions. As a result, M-band OWF designed by propose method has better performance than M-band OWF designed by exsiting method.

Block matching motion estimation is the most widely used method for motion compensated coding of image sequences. Based on a two dimensional systolic array, VLSI architecture and implementation of the full search block matching algorithm are described in this paper. The proposed architecture improves conventional array architecture by designing efficient processing elements that can control the data prodeuced by efficient search window division method. The advantages are that 1) it allows serial input to reduce pin counts for efficient composition of local memories but performs parallel processing. 2) It is flexible and can adjust to dimensional changes of search windows with simple control logic. 3) It has no idel time during the operation. 4) It can operate in real/time for low and main level in MPEG-2 standard. 5) It has modular and regular structure and thus is sutiable for VLSI implementation.

In this paper, we propose a new modification method of the look-up table in self-organizing fuzzy control using look-up table. This method has the property that look-up table is modified to have fast response property. Its principle is that the controller forces the trajectory to go into the fast respose region which the error change amount is larger than the error at initial time whenever the reference or disturbance change. Also we introduce the variable learning speed coefficient which is proportional to distance from switching curve. And to demonstrate the applicability of the proposed method, we had simulation study for some examples and esecuted pole balance experiments with inverted pendulum.

The elliptic filters are optimal in the sense that the magnitude characteristic exhibits the steepest slope at the cutoff frequency. The other characteristics such as delay and step response, however, are rather undesirable. In this paper, the modified elliptic low-pass filter function is proposed. The modified elliptic function possesses progressively diminishing ripples in both passband and stopband, and the lower pole-Q, values when compared to the elliptic counterparts, thus producting the flatter delay characteristics and improved time-domain performances. And it is realizable in the doubley-terminated ladder structures for the order n even or odd, thus lending themselves amaenable to high-quality active RC or switched capacitor filters through the simulation techniques.

Mathematical morphology based on set theory is easy to be implemented in parallel and can be applied to various fields in image analysis. Particularly mophological pattern spectrum can detect critical scales in an image object and quantify various aspects of the shape-size content. In this paper, texture classification using pattern spectrum based on morphological subband decomposition is porposed. The low-low band extracts pattern spectrum features, and the high-low, low-high, and high-high bands extrack the structural information. This approach has the advantages of efficient information extraction, less time-consuming, high accuacy, less computation, and parallel implementation.

This paper proposes a motion estimation method in video sequences based on the feature based matching and anistropic propagation. It measures translation and rotation parameters using a relaxation scheme at feature points and object orinted anistropic propagation in continuous and discontinuous regions. Also an iterative improvement motion extimation based on the strongly coupled module fusion and adaptive smoothing is proposed. Computer simulation results show the effectiveness of the proposed algorithm.

In this paper, we present a digital image data compression method using vector quantization preserving edges. A new vector quantization algorithm is proposed using a new sampling method and edge region extraction. The codebook generation time is faster than existing algorithms and the quality of decompressed images is much improved. Extrimental results suggest that the resultant compression ratio and PSNR are better than those of BPVQ and HMVQ methods.

This paper proposes a motphological fast reconstructive filter (FRF) using up/down sampling techniques for reconstructive opening and closing, and a parallel structure for fast multiresolution decomposition. Compuer simulation shows that, compared with the conventional RF, the proposed FRF can reduce the processing time up to 8 times while it maintains a similar performance in reconstructed shapes. Further reduction in the decomposition time achieved by the paralellized algorithm combined with the FRF, which can be applied in areas such as defect detection, image segmentation, pattern recognition, etc.

This paper proposes the Korean generator which is used in a Japanese to Korean translation system. The established Korean generator for japanese to Korean translation uses the entire sentence structure to generate Korean, or directly translates each morpheme. But in this paper, the generator generates Korean, based on a short clause. Japanese and Korean have similarities in short clauses, including the word order. But in the Korean generation, they also have many differences within short clauses. The analysed features in Japanese and Korean are in the wrong order. Therefore the Korean generator which takes it into consideration is needed. We propose a annotational algorithm for the Korean generation and apply it to the translation of postpositions, declinable words and auxiliary verbs. We also consider Korean phonological phenomenon that can't be extracted from the analyed Japanese informations. In the end, we can translate the Japanese conjugation that can't be easily dealt in the procedural method.

This article describes the hardware design scheme of Hamming MAXNET algorithm which is appropriate for binary pattern classification with minimum HD measurement between stimulus vector and storage vector. Circuit integration is profitable to Hamming MAXNET because the structure of hamming network have a few connection nodes over the similar neuro-algorithms. Designed hardware is the two-layered structure composed of hamming network and MAXNET which enable the characteristics of low power consumption and fast operation with biline volgate sensing scheme. Proposed Hamming MAXNET hardware was designed as quantize-level converter for simulation, resulting in the expected binary pattern convergence property.

Most pattern classifiers have been designed based on the ML (Maximum Likelihood) training algorithm which is simple and relatively powerful. The ML training is an efficient algorithm to individually estimate the model parameters of each class under the assumption that all class models in a classifier are statistically independent. That assumption, however, is not valid in many real situations, which degrades the performance of the classifier. In this paper, we propose a minimum-error-rate training algorithm based on the MAP (Maximum a Posteriori) approach. The algorithm regards the normalized outputs of the classifier as estimates of the a posteriori probability, and tries to maximize those estimates. According to Bayes decision theory, the proposed algorithm satisfies the condition of minimum-error-rate classificatin. We apply this algorithm to NPM (Neural Prediction Model) for speech recognition, and derive new disrminative training algorithms. Experimental results on ten Korean digits recognition have shown the reduction of 37.5% of the number of recognition errors.

A modified Hybrid learning rule(MHLR) is proposed, which is derived from combining the Back Propagation algorithm that is known as an excellent classifier with modified Hebbian by changing the orginal Hebbian which is a good feature extractor. The network architecture of MHLR is multi-layered neural network. The weights of MHLR are calculated from sum of the weight of BP and the weight of modified Hebbian between input layer and higgen layer and from the weight of BP between gidden layer and output layer. To evaluate the performance, BP, MHLR and the proposed Hybrid learning rule (HLR) are simulated by Monte Carlo method. As the result, MHLR is the best in recognition rate and HLR is the second. In learning speed, HLR and MHLR are much the same, while BP is relatively slow.

In this paper, we propose a new approach of solving the stereopsis problem with a discrete-time cellular neural network(DTCNN) where each node has connections only with its local neithbors. Because the matching process of stereo correspondence depends on its geometrically local characteristics, the DTCNN is suitable for the stereo correspondence. Moreover, it can be easily implemented in VLSI. Therefore, we employed a two-layer DTCNN with dual templates, which are determined with the back propagation learning rule. Based on evaluation of the proposed approach for several random-dot stereograms, its performance is better than that of the Marr-Poggio algorithm.

Presented is the study on design and implementation of a fuzzy system to approximately reason using measured actual storage rate in irrigation reservoirs. To design Fuzzy reasoning systems. Minimum Operation Rule by Mamdani was applied. Fuzzy variable and membership functions are determined after identifying storage-rate affecting factor and followed simulation. Hydrological model to express actual situation within drought areal boundary is generally too complex. Hereby, considering irregularity of time-rate storage change during irrigation, this system uses irrigation water and meteorological data as a IN-data. It was abvious the results were closely corresponding to the actual data observed.

In this paepr, we present a soft switching PWM inverter as reducing switching loss and stress at high power application. The PWM inverter is designed by space voltage about 2$\sqrt{3}$ times (15%) than conventional SPWM method. To reduce switching loss and stress. The DC link requires a capacitor, an inductor and two additional switches. Therefore proposed inverter performs trun PWM operation under the soft switching condition. As a result of soft switching we can reduce switching loss and ensure stability of switching devices. For approach to real time, control system is realized by 8 bit single-chip microprocessor. Therefore, we can construct system is with simplified volumn and structure by eliminating carrier wave and referrence wave generator of conventional SPWM method.