Among the various methods of realizing low sensitivity digital filters. LBR two-pair cascading is one which enables us to realize low sensitivity filter without referring to any analog prototype filters. In this paper we analyze the LBR cascade structures through the fundamental approach to them based on the WDF theory, and interpret their features described solely in the digital domain in relation to the corresponding analog circuit Besides, we derive in a unified manner various LBR constraints needed for the LBR cascade realization, point out the insufficiency lurking in the existing description of the Type 3 LBR constraints, complement it, and make the LBR two-pair cascading procedure complete.

For the conventional systme based on word, it is very difficult to enlarge the number of vocabulary. To cope with this problem, we must use more fundamental units of speech. For example, syllables and phonemes are such units, Korean speech consists of initial consonants, middle vowels and final consonants and has characteristic that we can obtain syllables from speech easily. In this paper, we show a speech recognition system with the advantage of the syllable characteristics peculiar to the Korean speech. The algorithm of recognition system is the Time Delay Neural Network. To recognize many recognition units, system consists of initial consonants, middle vowels, and final consonants recognition neural network. At first, our system recognizes initial consonants, middle vowels and final consonants. Then using this results, system recognizes isolated words. Through experiments, we got 85.12% recognition rate for 2735 data of initial consonants, 86.95% recognition rate for 3110 data of middle vowels, and 90.58% recognition rate for 1615 data of final consonants. And we got 71.2% recognition rate for 250 data of isolated words.

In this paper, we propose a probabilistic model appropriate for the bearing estimator which uses cross-correlation method following a close investigation on real underwater acoustic bearing data. The well-known JPDA(Joint Probabilistic Data Association) filter is tuned to the underwater acoustic bearing estimation based on the result that the reliability of the bearing measurement is related to the amplitude of the cross-correlation peak. The proposed probabilistic model is shown to be adequate by presenting the results of the improved tracking performance of the modified filter for various real bearing data as well as artificially generated ones.

The design and implementation of a fuzzy self-organizing controller for an environmental temperature chamber is discussed. The chamber is a non-linear, time-variant system with delay-time and dead-time. And the parameter tuning is required in PI control when the performance degraded. However the proposed fuzzy-SOC monitors the performance of the process. modifies the data base, and performs the delay-time compensation based on the idealized process model. A series of experiments was performed for the conventional PI and the fuzzy-SOC. These experimental results show the usefulness of the fuzzy-SOC.

The conventional resolved motion control is not applicable for the control of robots on the spacecraft on account of the floating base of the robots. When the robots perform the assembly or repair operations, the position and attitude of the base satellite are disturbed by the reaction force/moment caused by the robot motion. This reaction will cause error on the robot. motion. Therefore, we define a new type of Jacobian(Extended Jacobian) to minimize the effects of reaction on the accuracy of the robot performing assembly or repair operations. In this paper, we utilize the redundancy of the closed chain system to minimize the effects of the robot motion on the position and attitude of spacecraft. This will results in the accurate assembly and repair operations by the robot.

This paper describes the design principles and methods of electronic processor for thermal imager with the SPRITE detector, operating in the 8-12 micron band. The thermal imager consists of a optical scanner containing the detector and an electrical signal processor. The optical scanner utilizing rotating polygon and oscillating mirror, is 2-dimensional serial/parallel scan type using 5 elements of the detector. And the electronic processor has pre-processing of 5 chnanel's thermal signal from the detector, and performs digital scan conversion to reform the parallel data stream into serial analog data compatible with conventional RS-170 video. Through the designed electronic processor, we have acquired a satisfactory thermal image. And the MRTD (Minimum Resolvable Temperature Difference) is 0.5$^{\circ}$K at 7.5 cycles/mm.

The logic function synthesis using ULM U$_h$ is constructed based on canonic Reed-Muller expansion coefficient for a given function. This paper proposes the fast synthesis algorithm using ULM U$_h$ for ternary function. By using circuit cost and synthesis method of proposed in this paper, order of control input variable minimum number of ULM U$_h$ can be decided in the synthesis of n-variable ternary function. Accordingly, this method enables to optimum circuit realization for ternary function synthesis using ULM ULM U$_h$ and can be applied to ternary function synthesis using ULM U$_h$. The complexity of search for select the order of all control input variables is (n+2)(n-1)/2.

In this paper, we propose orthogonal integer transform(OIT) with general form. Considering the orthogonality and magnitude value order of the DCT Matrix whose performance is found to be close to that of the KLT, known to be optimal. The proposed OIT matrix is composed of values minimizing Hibert-Schmidt norm among integer values which satisfy the condition of orthogonality and the relative magnitudes of the DCT matrix. Comparing the OIT with the DCT, CMT, and ICT in error characteristics, transform efficiency, and maximum reducible bit, it is shown that the performance of the OIT compares more closely to that of the KLT relative to the performances of the DCT, CMT, and ICT when N=8.

In this paper, a new variable block-size transform coding algorithm using information of segmentation is proposed. In general variable block-size transform coding algorithms, the information which is used to form variable blocks is thrown away, but in the proposed algorithm, the information of segmentation which is composed of texture and contour information is maintained for efficient coding. That is, the number of blocks can be reduced by contour information, a DC component of block can be predicted by texture information, and the region compensation, which reduces mean difference between regions, using texture and contour information can diminish the magnitude of AC components. The result is that the proposed algorithm has improved highly the coding efficiency of DC and AC components.

This paper describes the recognition of the printed Hanguel(Korean Character) using Neural Network. In this study, Neural network is used in only specific classification. Hanguel is classified globally by using template matching. Neural network is learned using the segmented grapheme. The grapheme of Hanguel is segmented using the structural method. Neural network is constructed, which is corresponded to the kind and the shape of graphemes. Each neural network is multi layer perceptron. The learning algorithm is the modified error back propagation using descending epsilon method. With five test character sets, the recognition rate of 94.95% is obtained.

A new neural network architecture called the Parallel. Self-Organizing Hierarchical Neural Network (PSHNN) is presented. The new architecture involves a number of stages in which each stage can be a particular neural network (SNN). The experiments performed in comparison to multi-layered network with backpropagation training and indicated the superiority of the new architecture in the sense of classification accuracy, training time,parallelism.

In this paper, we develop an algorithm which recognizes Ilpa-style shorthand characters by on-line. It discriminates the structure of characters using coordinates which are measured by tablet board, then it outputs the recognized characters using the fuzzy inference rules. Shorthand characters have several forms, in which an initial or a middle sound depends on angle and length while a last sound is treated as a hook. We apply fuzzy inference rules to the discrimination of the length, the angle, the curve, and the straight line. We also built up a set of standard character codes in order to reduce the processing time.

Hardware implementation of the on-chip learning artificial neural networks is important for real-time processing. A pRAM model is based on probabilistic firing of a biological neuron and can be implemented in the VLSI circuit with learning capability. We derive a backpropagation learning algorithm for the pRAM networks and present its circuit implementation with stochastic computation. The simulation results confirm the good convergence of the learning algorithm for the pRAM networks.