• Geophysics and Geophysical Exploration
• /
• v.1 no.3
• /
• pp.155-160
• /
• 1998

EM tomography technique has been developed. The algorithm used the extended Born approximations for forward modeling and reconstructed a conductivity image by a smoothness constraint least squares inversion method. Observed data, the vertical components of secondary magnetic fields, were simulated with the 3-D integral equation code. The results showed that the location of anomalous body could be imaged very well, but conductivity of the body was lower than real one and the vertical resolution was much higher than the horizontal resolution.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.6 no.2
• /
• pp.28-33
• /
• 1988

The objective of this work is to study the sue of precise triangulation nets by matrix analysis in adjusting coordinates. Results of this work show that matrix analysis has many advantages, namely in terms of accuracy and efficiency. In addition to this, comparison of adjusted vlues by two different methods, a Rigorous Method and an Approximate Method, show that differences in adjusted coordinates are 1-2 mm within the mean square error. This proves that the Approximate Method has sufficient accuracy in the practical application.

• Korean Journal of Optics and Photonics
• /
• v.13 no.4
• /
• pp.275-282
• /
• 2002

We derived analytic formulae for the correction of spherical aberration, coma, and axial color of a two-components lens system consisted of a cemented doublet and a singlet by using the thin lens approximation. The correction formulae were applied to design a telephoto lens system. We examined two kinds of glass combinations in the design, one was crown-flint-crown combination and the other was flint-crown-flint combination. We found two kinds of achromatic aplanat solutions in the crown-flint-crown combination. For the case of flint-crown-flint combination, there were also two kinds of solutions, but their configurations are not useful in practice.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.3 no.3
• /
• pp.597-605
• /
• 1999

In this paper, we propose a cardioangiography sequence image coding scheme which use a subtraction between initial image and current frame inserted contrast dye. Stable regions are obtained by the multithreshold and meaningful region is extracted by the images with stable region. The image with meaningful region is classified into contour and texture information. Contour information is coded by contour coding. And texture information is approximated by two-dimensional polynomial function and each coefficients is coded. Experimental results confirm that the sequence of cardioangiography are well reconstructed at the low bit rate (0.02∼0.04 bpp) and high compression ratio.

• Geophysics and Geophysical Exploration
• /
• v.2 no.2
• /
• pp.86-95
• /
• 1999

This paper presents an efficient three-dimensional (3-D) modeling algorithm using the extended approximation to an electric field integral equation. Numerical evaluations of Green's tensor integral are performed in the spatial wavenumber domain. This approach makes it possible to reduce computing time, to handle smoothly varying conductivity model and to remove singularity problems encountered in the integration of Green's tensor at a source point. The responses obtained by 3-D modeling algorithm developed in this study are compared with those by the full integral equation for a thin-sheet EM scattering. The extensive analyses on the performance of modeling algorithm are made with the conductivity contrasts and source frequencies. These results show that the modeling algorithm are accurate up to the conductivity contrast of 1:16 and the frequency range of 100 Hz-100 kHz. The extended Born approximation, however, may produce inaccurate results for some source and model configurations in which the electric field is discontinuous across the conductivity boundary. We performed the modeling of a composite model of which conductivity varies continuously and this shows the modeling algorithm developed in this study is efficient for 3-D EM modeling. For a cross-hole source-receiver configuration a composite model of which conductivity varies continuously can be successfully simulated using this algorithm.

• Proceedings of the IEEK Conference
• /
• 2003.07a
• /
• pp.218-221
• /
• 2003

본 논문에서는 사전등화를 이용하는 상향링크 MC-CDMA(multicarrier-code division multiple access)/TDD(time division duplexing) 시스템에서 사전등화를 위한 상향링크 채널을 추정하는 방법들을 제안하고 시스템의 성능분석을 수행한다. 제안된 방법들에서는 하향링크 슬롯구간에서의 채널변화를 적절한 차수의 다항식으로 모델링하고, 이 다항식을 상향링크 슬롯구간으로 확장함으로써 상향링크 슬롯구간의 채널을 추정한다. 하향링크 슬롯구간에서의 채널변화는 MMSE(minimum mean squared error)curve fitting 방법이나 Lagrange 보간법 등이 사용되며 1차, 2차, 3차 다항식으로 근사화 된다. 성능지표로 정확도보다 시스템 성능이 중요 하므로 BER (bit error rate)을 사용한다. 다양한 시스템 및 채널환경에서의 모의실험 결과로부터 Lagrange 보간법은 하향링크 채널정보가 정확한 경우에는 MMSE 방법보다 성능이 다소 우수하지만 하향링크 채널추정 오류에 매우 민감하며, 2 차 다항식을 사용한 MMSE curve fitting 방법은 다양한 환경에서 우수한 성능을 가질 뿐만 아니라 채널추정 오류에도 매우 강인함을 알 수 있다.

• The Journal of the Korea Contents Association
• /
• v.21 no.3
• /
• pp.616-625
• /
• 2021

Deep neural networks are an approximation method that approximates an arbitrary function to a linear model and then repeats additional approximation using a nonlinear active function. In this process, the method of evaluating the performance of approximation uses the loss function. Existing in-depth learning methods implement approximation that takes into account loss functions in the linear approximation process, but non-linear approximation phases that use active functions use non-linear transformation that is not related to reduction of loss functions of loss. This study proposes parametric activation functions that introduce scale parameters that can change the scale of activation functions and location parameters that can change the location of activation functions. By introducing parametric activation functions based on scale and location parameters, the performance of nonlinear approximation using activation functions can be improved. The scale and location parameters in each hidden layer can improve the performance of the deep neural network by determining parameters that minimize the loss function value through the learning process using the primary differential coefficient of the loss function for the parameters in the backpropagation. Through MNIST classification problems and XOR problems, parametric activation functions have been found to have superior performance over existing activation functions.

• Proceedings of the Korea Contents Association Conference
• /
• 2005.05a
• /
• pp.341-345
• /
• 2005

In this paper, we propose an algorithm for the approximation of FIR filters by IIR filters. The algorithm is based on a concept of the balanced model reduction. Head-related transfer functions(HRTFs) of dummy-head are approximated by 32-order IIR filters. The binaural sounds using the approximated HRTFs are reproduced by headphone, and serves as a cue of sound image localization. Experiment of sound image are carried out for 10 participants with computer simulation and DSP board respectively. The results of the experiments show that the localization using the approximated HRTFs by IIR filters is the same accuracy as the case of FIR filters that simulate the HRTFs.

• Computational Structural Engineering
• /
• v.11 no.1
• /
• pp.171-178
• /
• 1998

The added viscoelastic dampers increase damping and stiffness of buildings and results in so called non-classical or non-proportional damping problem. In this system the eigenvectors of the undamped system may not diagonalize the damping matrix, and the system is generally analyzed by converting the equation of motion into a 2n first order state-space form. As this approach is complex and time-consuming compared to the classically damped problem, the system is often analyzed by neglecting the off-diagonal terms in the damping matrix. In this paper the theoretical background of the approximate approach is studied, and the vibration characteristics of a three-story shear building with a viscoelastic damper are investigated using the exact and approximate method. It is found that the approximate method may produce good result when the additional damping is small, but as the damping increases the error also increase.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.39A no.6
• /
• pp.322-332
• /
• 2014

This paper proposes a robust lane detection algorithm for non-flat roads by combining a piecewise linear model and dynamic programming. Compared with other lane models, the piecewise linear model can represent 3D shapes of roads from the scenes acquired by monocular camera since it can form a curved surface through a set of planar road. To represent the real road, the planar roads are created by various angles and positions at each section. And dynamic programming determines an optimal combination of planar roads based on lane properties. Experiment results demonstrate the robustness of proposed algorithm against non-flat road, curved road, and camera vibration.