Although the fuzzy logic controller (FLC) has been adopted in many engineering applications, one hesitates to adopt the FLC in critical applications, since there was no definite control theoretic analysis. In this paper, based on the stability/robustness analysis of an FLC by S.Y.Yi$^{[3]}$, we apply the FLC to robot manipulator with the structured and unstructured uncertainties e.g., load variation and firction, etc. And we verify the performance of the FLC by computer simulation on a simple two-link robot manipulator.

This paper presents a finger-shaped multisensor system which can measure the tyep and position of a target surface by contactl. The multi-sensor system consists of a sphere-shpaed optical tactile sensor located at the finger tip and a force/torque sensor located at the joint of a finger. The optial tactile sensor determines the type and position of the target surface using the shape and position of the CCD image of the touching area generated by a contact between the sensor and the taget surface. The force/torque sensor also determines the position and surface normal vector by applying the distributionof forces and torques t the contact point to the equations of finger shape. The measurements on the position and surface normal vector at a contact point obtined by two individual sensors are fused using a statistical method. The integrated sensor system has 0.8mm error in position measurement and 1.31$^{\circ}$ error in normal vector measurement. The developed sensor system has many applications, such as autonomous compliance control, automatic grasping and recognition, etc.

In this paper, we propose a coprime factor model reduction method to get a reduced order controller in $H^{\infty}$ mixed sensitivity problem with frequency weighting functions. for this purpose, the given $H^{\infty}$ mixed sensitivity problem is transformed into robust stabilization problem with coprime factor uncertainty of given plant. This method is to define frequency weighted coprime factors of plant in CSD (chain scattering description) form and reduce the coprime factors using weighted balanced truncation. then a controller is designed to the reduced order coprime factors using J-lossless coprime factorization method. Using this approach, the robust stability condition is derived and good performance is preserved in closed loop system with the given plant and the reduced order controller. Also the order of reduced controller for guaranteeing the robust stability can be determined before designing the reduced controller. The proposed method behaves well in both stable and unstable plant.

By utilizing the interinsic structure of the underlying continous-time nonlinear system, one can design an approximate sampled-data observer improved with respect ot the sampling-time for the systems. In this paper, we characterize the conditions for the solvability of the improved approximate sampled-data nonlinear observer design problem. In particular, it is shown that when the dimension of the state space is two, the nonlinear systems for which it is possible ot desing 3rd or higher order approximate sampled-data nonlinear observer are locally state-equivalent to an observable bilinear system. The practical implication is that seeking higher order approximate sampled-data nonlinear observer for nonlinear systems is very restricted.

This paper proposes a cost-effective solution to the synonym problem of virtual caches. In the proposed solution, a minimal hardware addition guarantees the correctness whereas the software counterpart helps improve the performance. The key to this proposed solution is an addition of a small physically-indexed cache called U-cache. The U-cache maintains the reverse translation information of the cache blocks that belong to unaligned virtual pages only, where aligned measns that the lower bits of the virtual page number match those of the corresponding physical page number. The page alignment is a simple software optimization to improve the performance of the U-cche hardware. With the combination of both hardware and software, the proposed solution reduces the hardware costs and minimizes software modification and performance degradation. Performance evaluation base on ATUM traces shows that a U-cache, with only a few entries, performs almost as well as fully-configured hardware-based solution when more than 95% of the pages are aligned.

A microwave incoherent imaging method for a conducting cyliner by using multi-frequency tiem-harmonic field is presented in this study. In this paper, an incoherent intensity pattern of th econducting cylinder is obtained by averagin gout the multi-frequency intensities of the coherent field such as the time-harmonic field scattered from this cylinder. This phenomenon is hsown numerically in scattering by a conducting circular cylinder illuminated by the time-harmonic plane wave, and is interpreted analytically by the mutual coherence functon defined as a frequency-averaged intensity of the time-harmonic fields in th frequency domain.

In the previous paper of this study, an incoherent intensity pattern of a conducting cylinder is obtained by averaging out the multi-frequency intensities of the time-harmonic fields scattered from this cylinder. in this paper, an incoherent imaging of a conducting cylinder is obtained via the back-projection of the incoherent intensity patterns calculated from the circular rotational measurement configuration. This imaging method is vlaidated by imaging a conducting circular and a conducting elliptic cylinder and the measurement conditions to get better image are investigated.

This paper describes the issue relating to the design and implementation of a high-precision vision system for monitor inspection. We present several monitor inspection algorithms which evaluate the geometric distrotions of the patterns on the inspection target monitor. Also, to inspect color distortion acquired form color CCD camera into the XYy coordinate, accurately. The proposed monitor inspection system is implemented on IBM PC/486, and the user interface is developed on microsoft windows to assist unskilled users for easy use. Using the developed vision system, Experiments on the inspection accuracy and speed were performed and the results are presented.

Existing area-based stereo matching algorithms utilize a single rectangular correlation area for computing cross-correlation between corresponding points in stereo images, and compute disparity by finding the peak in the vicinity of depth discontinuity, since, because of inconstnat disparities around discontinuities, the cross-correlation becomes low in such area. Inthis paper, a new area-based matching strategy is proposed exploiting multiple directional correlation masks instead of a single one. The proposed technique computes multiple cross-covariance functions using each oriented mask. Peaks are detected from each covariance function and the disparity is computed by choosing the location with the highest covariance value. Proposed approach can also be applied to compute disparity gradients without obtaining dense depth data. A number of examples are presented using synthetic and natural stereo images.

In this paper, we present a half rate speech coder using trellis excitation. The coder combines code-excited linear prediction (CELP) system and trellis quantization method using the codebook expansion, and it produces higher speech quality than the typical CELP coder for the same transmission rate. A subjective comparison with 3~8 bit .$\mu$-law PCM indicates that the half rate coder provides speech quality between 5-bit and 6-bit $\mu$-law PCM .

This paper describes the realization of a rela-time adaptive acoustic echo canceller, which adopts a microprogramming method, for removing acoustical echoes in speakerphone systems using th eADSP-2101 microprocessor with a pipeline and modified harvard architecture. We apply the LMS (least mean square) algorithm to estimate the coefficients of a transversal FIR filter. For the acustic adaptive echo canceller, we propose a parallel operation programming to imrove algorithm execution speed and apply a nonlinear quantization to reduce the quantization error caused by large dynamic range of voice signal.

In this paper, we propose a unified systolic array for the computation of the 2D discrete cosine transform/discrete sine transform/discrete hartley transform (DCT/DST/DHT). The unified systeolic array for the 2D DCT/DST/DHT is a generalization of the unified systolic array for the 1D DCT/DST/DHT. In order to calculate the 2D transform, we compute 1D transforms along the row, transpose them, and obtain 1D transforms along the column. When we compare the proposed systolic array with the conventional method, our architecture exhibits a lot of advantages in terms of latency, throughput, and the number of PE's. The simulation results using very high speed integrated circuit hardware description language (VHDL), international standard language for hardware description, show the functional validity of the proposed architecture.

In tis paper, a new morphological algorihtm for binary image reconstruction from graayscale skeleton is proposed. In the proposed algorithm, grayscale morphological dilation is utilizedm instead of binary morphological operatons for the conventional methods. The algorithm is proven mathematically and verified by computer simulation results.

Fuzzy inference is very useful in expressing ambiguous problems quantitatively and solving them. But like the most of the knowledge based inference systems. It has many difficulties in constructing rules and no learning capability is available. In this paper, we proposed a fuzzy inference system based on fuzy associative memory to solve such problems. The inference system proposed in this paper is mainly composed of learning phase and inference phase. In the learning phase, the system initializes it's basic structure by determining fuzzy membership functions, and constructs fuzzy rules in the form of weights using learning function of fuzzy associative memory. In the inference phase, the system conducts actual inference using the constructed fuzzy rules. We applied the fuzzy inference system proposed in this paper to a pattern classification problem and show the results in the experiment.

This paper describes the chaotic neuron model to represent the complicated states of brain and analyzes the dynamical responses of chaotic neuron such as periodic, bifurcation, and chaotic phenomena which are simulated iwth numerical analysis. Next, the chaotic neuron circuit is implemented w ith the analog electronic devices. The transfer function of chaotic neuron is given by summed the linear and nonlinear property. The output function of chaojtic neuron is designed iwth the two cMOS inverters and a feedback resistor. By adjusting the external voltage, the various dynamical properties are demonstrated. In addition, we construt the chaotic neural networks which are composed of the interconnection of chaotic neuroncircuit such as serial, paralle, and layer connection. On the board experiment, we proved the dynamci and chaotic responses which exist in the human brain.

This paper proposes a combined feedforward/feedback controller which uses jacobians of neural network. The jacobians are calculated form the neural network that identifies the nonlinear plant, which are used for designing a jacobian controller and for training a neural network controller. Normally, it takes much time to train the neural network controller. Combining the neural and the jacobian controller, it can be a stable controller from the beginning of training phase of neural network, and it can be implemented as a learning-while-functioning controller. Simulated resutls for the proposed controller show its effectiveness and better performances.

In this paper we describe designing and implementing a digital neural chip and a parallel neural machine for simulating large scale neural netsorks. The chip is a single-chip multiprocessor which has four digiral neural processors (DNP-II) of the same architecture. Each DNP-II has program memory and data memory, and the chip operates in MIMD (multi-instruction, multi-data) parallel processor. The DNP-II has the instruction set tailored to neural computation. Which can be sed to effectively simulate various neural network models including on-chip learning. The DNP-II facilitates four-way data-driven communication supporting the extensibility of parallel systems. The parallel neural machine consists of a host computer, processor boards, a buffer board and an interface board. Each processor board consists of 8*8 array of DNP-II(equivalently 2*2 neural chips). Each processor board acn be built including linear array, 2-D mesh and 2-D torus. This flexibility supports efficiency of mapping from neural network models into parallel strucgure. The neural system accomplishes the performance of maximum 40 GCPS(giga connection per second) with 16 processor boards.

In this paper, a new competitive learning algorithm called dynamic competitive learning (DCL) is presented. DCL is a supervised learning mehtod that dynamically generates output neuraons and nitializes weight vectors from training patterns. It introduces a new parameter called LOG (limit of garde) to decide whether or not an output neuron is created. In other words, if there exist some neurons in the province of LOG that classify the input vector correctly, then DCL adjusts the weight vector for the neuraon which has the minimum grade. Otherwise, it produces a new output neuron using the given input vector. It is largely learning is not limited only to the winner and the output neurons are dynamically generated int he trining process. In addition, the proposed algorithm has a small number of parameters. Which are easy to be determined and applied to the real problems. Experimental results for patterns recognition of remote sensing data and handwritten numeral data indicate the superiority of dCL in comparison to the conventional competitive learning methods.

In this paper, we propose an on-line recognition system of chinese characters using an adaptive resonance theory-1(ART-1) neural network. Strokes, primitive components of chinese characters are usually warped into a cursive form and classifying them is very difficult. To deal with such cursive strokes, we use an ART-1 neural network that has the following advantages: (1) it automatically assembles similar patterns together to form classes in a self-organized manner: (2) it directly accesses the recognition codes corresponding to binary input patterns after self-stabilizing; (3) it doesn't tends to get trapped in local minima, or globally incorrect solutions. A database for character recognition also dynamically constructed with generalized character lists, and a new character can be included simply by adding a new sequence to the list. Character recognition is achieved by traversing the chinese datbase with a sequence of recognized strokes and positional relations between the strokes. To verify the performance of the system. We tested it for 1800 daily-used basic chinese second per character. This results suggest that the proposed system is pertinent to be put into practical use.

This paper proposes a new design technique that can be used for BSB (brain-state-in-a-box) neural networks to realize autoassociative memories. The proposed method is based on the parametrization of solution space and optimization using genetic algorithm. The applicability of the established technique is demonstrated by means of a simulation example, which illustrates its strengths.

This paper introduces the concept of fixed weights and proposes an algorithm for classification by adding this concept to vector space separation method in LVQ. The proposed algorithm is based on competitive learning. It uses fixed weightsfor generality and fast adaptation efficient radius for new weight creation, and L1 distance for fast calcualtion. It can be applied to many fields requiring adaptive learning with the support of generality, real-tiem processing and sufficient training effect using smaller data set. Recognition rate of over 98% for the train set and 94% for the test set was obtained by applying the suggested algorithm to on-line handwritten recognition.

A neural network-based PSRI(position, scale, and rotation invariant) feature extraction and ATR (automatic target recognition) system are proposed and an efficient parallel implementatio of the proposed system using multicomputer is also presented. In the proposed system, the scale and rotationinvariant features are extracted from the contour projection of the number of edge pixels on each of the concentric circles, which is input t the cooperative network. We proposed how to decide the optimum depth and the width of the parallel pipeline system for real time applications by modeling the proposed system into a parallel pipeline implementation method using transputers is also proposed. The implementation results show that we can extract PSRI features less sensitive to input variations, and the speedup of the proposed ATR system is about 7.55 for the various rotated and scaled targets using 8-node transputer system.

This paper deals with the speed control of the switched reluctnace motor using fuzzy PI controller. A fuzzy logic control provides a good approach to nonlinear system because it does not require a detailed mathematical model to formulate the algorithm. The fuzzy PI controller is implemented by MCS80C196KB, a 16 bit one-chip microcontroller, and an EPROM is used for the commutation logic of the SRM. The simulation and experimental results show that the performance of the fuzzy PI controller is superior to that of the conventional PI controller in terms of response time, settling time and overshoot. In particular, the robustness of the system is largely improved.

Switched reluctance motor (SRM) drives appear today as viable alternatives to induction and brushless motor drives in a wide range of application such as machine tools, fans, and pumps. This paper deals with a new converter topology which can be used in order converter. The split source type converter consists of minimum switching devices to be used in switched reluctance motor (SRM) drive. In this proposed converter two switches and six diodes are added to split source type converter. The proposed converter has performance minimizes the negative torque, Which puts the phase current off by double impressed voltages. The major advangtage of this converter is the increase of the average output power while improving better converter efficiency in heavy load and high speed than other topologies. The proposed converter system has been implemented using 80C196KB microcontroller and experiments were carried out ot verify the simulation results.