An adaptive hierarchical classifier(AHCL) for Korean character recognition using a neural net is designed. This classifier has two neural nets: USACL (Unsupervised Adaptive Classifier) and SACL (Supervised Adaptive Classifier). USACL has the input layer and the output layer. The input layer and the output layer are fully connected. The nodes in the output layer are generated by the unsupervised and nearest neighbor learning rule during learning. SACL has the input layer, the hidden layer and the output layer. The input layer and the hidden layer arefully connected, and the hidden layer and the output layer are partially connected. The nodes in the SACL are generated by the supervised and nearest neighbor learning rule during learning. USACL has pre-attentive effect, which perform partial search instead of full search during SACL classification to enhance processing speed. The input of USACL and SACL is a directional edge feature with a directional receptive field. In order to test the performance of the AHCL, various multi-font printed Hangul characters are used in learning and testing, and its processing its speed and and classification rate are compared with the conventional LVQ(Learning Vector Quantizer) which has the nearest neighbor learning rule.

The paper proposes a design of high-speed dynamic CMOS PLA (Programmable Logic Array) which performs stable circuit operation. The race problem which nay occur in a NOR-NOR implementation of PLA is free in the proposed dynamic CMOS PLA by delaying time between the clocks to the AND- and to the OR-planes. The delay element has the same structure as the product line of the longest delay in the AND p`ane. Therefore it is unnecessary to design the delay element or to calculate correct delay time. The correct delay generated by the delay element makes the dynamic CMOS PLA to perform correct and stable circuit operation. Theproposed dynamic CMOS PLA has few variation of switching delay with the increasing number of inputs or outputs in PLA. It is verified by SPICE circuit simulation that the proposed dynamic CMOS PLA has the better performance over existing dynamic CMOS PLA's.

In this paper, multi pulse excited-linear predictive coding (MPE-LPC), where the correlation eliminated residual signal is modeled by a few pules, is shown to be effective for the compression of electrocardiogram (ECG) data, and a more efficient scheme for a faithful reconstruction of ECG is proposed. The reconstruction charateristic of QRS's and P.T waves is improved using the adaptive pulse allocation (APA), and the compression ratio (CR) can be changed by controlling the mumber of modeling pulses. The performance of the proposed method was evaluated using 10 normal and 10 abnormal ECG data. The proposed method had a better performance than the variable threshold amplitude zone time epoch coding (AZTEC) algorithm and the scan-along polygonal approximation (SAPA) algorithm with the same CR. With the CR in kthe range of 8:1 to 14:1, we could compress ECG data efficiently.

This paper proposes an on-line method of understanding humans conducting action observed through a vision sensor. The vision system captures images of conducting action and extracts image coordinates of endpoint of the baton. A proposed algorithm based on the expert knowledge about conducting recognizes patterns of the conducting action from the extracted image coordimates and play the corresponding music score. Complementary algorithms are also proposed for identifying the first beat static point and dynamics through extensive experiments, this algorithm is found to detect lower edges and upper edges without error.

In this paper a new turning gait is proposed for a quadruped walking robot. The proposed scheme makes it possible to control the translation and orientation of the walking robot simultaneously. At first the feasible leg sequences which can guarantee a positive longitudinal gait stability margin for each direction of movement are found. A method for finding the lifting time of each leg of a feasible leg sequince and selecting an optimal gait among feasible gaits is then suggested. The proposed gait can be appled to control the posture of walking robots and to generate an optimal gait for a desired movement of translation and rotation of the walking robot systematically.

In this paper we propose a hybrid position/force controller of a robot manipulator using feedback error learning rule and neural networks. The neural network is constructed from inverse dynamics. The weighting value of each neuron is trained by using a feedback force as an error signal. If the neural networks are sufficiently trained well, it does not require the feedback-loop with error signals. The effectiveness of the proposed hybrid position/force controller is demonstrated by computer simulation using PUMA 560 manipulator.

Automatic recognition of seal imprint has been required in the oriental countries. In this paper, a hierarchical approach for seal imprint verification is presented. Global features are used for seal imprint description in the first step. In the second step, conventional and several proposed local features are used to detect useful informations such as size, distribution and relative position of stroke length from seal imprint. In the last step, seal imprints are classified into one of three categories 'accept', 'ambiguous' and reject', based on the hierarchical classification. Experimental results show good performance on classification and recognition.

In this paper, we propose a motion estimation algorithm using hierarchical structure. The algorithm uses the image pyramids from the repetitive application of Gaussian filtering and decimation, and performs an inter-level displacement propagation in its motion estimation process. The motion estimation algorithm based on the hierarchical structure is shown to be very effective since this scheme utilizes the local imformation as well as the global imformation. The experimental results on the various data imdicate that compared to the Horn and Schunck's method, the proposed algorithm yields an accurate motion estimation with a fast convergence behaviour.

Conventional stereo matching techniques are based on the assumption that the features representing an object in left and right images have fixed attribute values. But, in fact, such features may take different values due to the practical stereo image formation and the image acquisition error, and thus the conventional techniques tend to result in the in the incorrect matching of features. In this paper, we propose a stereo matching mathod with a possibilistic view which copes with the possible variability of feature values. As a result, this method decreases the number of incorrect matching features when the values of corresponding features are somewhat large. The effectiveness of the proposed method is shown via computer simulation.

The multi-layer neural network that has broadly been utilized in designing the controller of robot manipulator possesses the desirable characteristics of learning capacity, by which the uncertain variation of the dynamic parameters of robot can be handled adaptively, and parallel distributed processing that makes it possible to control on real-time. However the error back propagation algorithm that has been utilized popularly in the learning of the multi-layer neural network has the problem of its slow convergencs speed. In this paper, an approach to improve the convergence speed is proposed using fuzzy logic that can effectively handle the uncertain and fuzzy informations by linguistic level. The effectiveness of the proposed algorithm is demonstrated by computer simulation of PUMA 560 robot manipulator.

The characteristics of a power plant changes as it operates for a long time and/or for different operating point. As a result, operators must retune gains of the controllers for better performance. In fact, skilled operators can retune the gains in reference to recorded data obtained by a test called dynamic test. The dynamic test, however, requires much time, and can be heavy burden for operators. In this paper, an expert knowledge-based auto-tuner is designed for drum-type boiler controllers of a fossil power plant using fuzzy logic. The performance of the proposed auto-tuner is shown via computer simulation and the simulation results show that the proposed auto-tuner is satisfactory for the desired performance.