• Journal of the Semiconductor & Display Technology
• /
• v.19 no.2
• /
• pp.100-104
• /
• 2020

Over the past several years, deep learning has been widely used for feature extraction in image and video for various applications such as object classification and facial recognition. This paper introduces an implantation of embedded Linux system for embossed digits recognition using CNN based deep learning methods. For this purpose, we implemented a coin recognition system based on deep learning with the Keras open source library on Raspberry PI. The performance evaluation has been made with the success rate of coin classification using the images captured with ultra-wide angle camera on Raspberry PI. The simulation result shows 98% of the success rate on average.

• Journal of the Korean Society for Precision Engineering
• /
• v.20 no.8
• /
• pp.175-184
• /
• 2003

A method of the shape design sensitivity analysis based on the boundary integral equation formulation is presented for two-dimensional inhomogeneous thermal conducting solids with multiple domains. Shape variation of the external and interface boundary is considered. A sensitivity formula of a general performance functional is derived by taking the material derivative to the boundary integral identity and by introducing an adjoint system. In numerical analysis, state variables of the primal and adjoint systems are solved by the boundary element method using quadratic elements. Two numerical examples of a compound cylinder and a thermal diffuser are taken to show implementation of the shape design sensitivity analysis. Accuracy of the present method is verified by comparing analyzed sensitivities with those by the finite difference. As application to the shape optimization, an optimal shape of the thermal diffuser is found by incorporating the sensitivity analysis algorithm in an optimization program.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2002.12a
• /
• pp.179-182
• /
• 2002

In this paper, this system makes use of the analog sensor and converts the feature of fish analog signal when sensor is operating with CPU(80C196KC). Then, After signal processing, this feature Is classified a special feature and a outline of fish by using the neural network, one of the artificial intelligence scheme. This neural network classifies fish pattern of very simple and short calculation. This has linear activation function and the error backpropagation is used as a learning algorithm. And the neural network is learned in off-line process. Because an adaptation period of neural network is too long time when random initial weights are used, off-line learning Is induced to decrease the Progress time We confirmed this method has better performance than somewhat outdated machines.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2002.12a
• /
• pp.410-413
• /
• 2002

일반적으로 휴대폰에는 리튬이온(Ll-lon) 전지(battery)를 많이 사용하고 있으며 그 전지(battery)를 충전시키기 위해 Microcontroller를 사용해서 과충전과 방전, 그리고 전지(battery) 보호와 충전에 대한 일정한 전류를 제어한다. 여기에서 충전 동작 시 필요한 일반직인 충전 전류 제어를 PWM의 방식에 의존하지 않고 인공지능 기법을 이용해 소프트웨어적으로 처리가 필요한 파라메터 값을 추정해 적용시키고자 한다. 따라서 개발한 충전시스템에 일반적인 충전 파라메터를 전압과 전류 그리고 시간으로 분류하여 Microcontroller에 그 파라메터를 적용시켜 PWM 방식으로 제어한 후에 실험에 의한 결과값을 얻는다. 그리고 이것들을 비교하여 보다 나은 충전시스템을 구현하기 위해 인공지능 기법 중에 하나인 신경망을 이용하여 전압과 전류 그리고 시간에 대한 파라메터를 처리하였다. 본 논문에서 신경망에 대한 파라메터의 학습을 일반 FC에서 구현하고 여기에서 추출된 학습 값을 Microcontroller에 적용시켜 입력값에 따라 다양한 PWM 신호를 발생시키도록 구현했다. 이후 실제적인 실험에 의한 결과값을 본 논문에서 서술하였다.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.9 no.4
• /
• pp.309-314
• /
• 2009

This paper presents implementation of the adaptive neuro-fuzzy control method. Control performance of the adaptive neuro-fuzzy control method for a popular inverted pendulum system is evaluated. The inverted pendulum system is designed and built as an education kit for educational purpose for engineering students. The educational kit is specially used for intelligent control education. Control purpose is to satisfy balancing angle and desired trajectory tracking performance. The adaptive neuro-fuzzy controller has the Takagi-Sugeno(T-S) fuzzy structure. Back-propagation algorithm is used for updating weights in the fuzzy control. Control performances of the inverted pendulum system by PID control method and the adaptive neuro-fuzzy control method are compared. Control hardware of a DSP 2812 board is used to achieve the real-time control performance. Experimental studies are conducted to show successful control performances of the inverted pendulum system by the adaptive neuro-fuzzy control method.

• Journal of Communications and Networks
• /
• v.18 no.3
• /
• pp.446-459
• /
• 2016

In this paper, we propose virtual direction multicast (VDM) for video multicast applications on peer-to-peer overlay networks. It locates the end hosts relative to each other based on a virtualized orientation scheme using real-time measurements. It builds multicast tree by connecting the nodes, which are estimated to be in the same virtual direction. By using the concept of directionality, we target to use minimal resources in the underlying network while satisfying users' quality expectations. We compare VDM against host multicast tree protocol.We simulated the protocol in a network simulator and implemented in PlanetLab. Results both from simulation and PlanetLab implementation show that our proposed technique exhibits good performance in terms of defined metrics.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.39 no.3
• /
• pp.223-231
• /
• 1990

Optimal Power Flow (OPF) solution by the Newton's method provides a reliable and robust method to classical OPF problems. The major challenge in algorithm development is to identify the binding inequalities efficiently. This paper proposes a simple strategy to identify the binding set. From the mechanism of penalty shifting with soft penalty in trial iteration, an active binding set is identidied automatically. This paper also suggests a technique to solve the linear system whose coefficients are presented in the matrix from. This implementation is highly efficient for sparsity programming. Case studies for 3, 5, 14, 118 bus and practical TPC-190, KEPCO-306 bus systems are performed as well.

• Journal of Electrical Engineering and information Science
• /
• v.2 no.5
• /
• pp.79-88
• /
• 1997

This paper introduces a new energy function, as maximum a posteriori(MAP) estimate of binocular disparity, that can deal with both random dot stereo-gram(RDS) and natural scenes. The energy function uses phase-magnitude as features to detect only the shift for a pair of corrupted conjugate images. Also we adopted Fleet singularity that effectively detects unstable areas of image plant and thus eliminates in advance error-prone stereo mathcing. The multi-scale concept is applied to the multi laser architecture that can search the solutions systematically from coarse to fine details and thereby avoids drastically the local minima. Using mean field approximation, we obtained a compact representation that is suitable for fast computation. In this manner, the energy function satisfies major natural constraints and requirements for implementing parallel relaxation. As an experiment, the proposed algorithm is applied to RDS and natural stereo images. As a result we will see that it reveals good performance in terms of recognition errors, parallel implementation, and noise characteristics.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.52 no.5
• /
• pp.295-304
• /
• 2003

In this paper, we proposed a hardware-oriented Genetic Algorithm Processor(GAP) based on subpopulation architecture for high-performance convergence and reducing computation time. The proposed architecture was applied to enhancing population diversity for correspondence to premature convergence. In addition, the crossover operator selection and linear ranking subpop selection were newly employed for efficient exploration. As stochastic search space selection through linear ranking and suitable genetic operator selection with respect to the convergence state of each subpopulation was used, the elapsed time of searching optimal solution was shortened. In the experiments, the computation speed was increased by over $10\%$ compared to survival-based GA and Modified-tournament GA. Especially, increased by over $20\%$ in the multi-modal function. The proposed Subpop GA processor was implemented on FPGA device APEX EP20K600EBC652-3 of AGENT 2000 design kit.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.11
• /
• pp.1424-1428
• /
• 1999

This paper presents a mathematical decomposition coordination method to implementing the distributed optimal power flow (OPF), wherein a regional decomposition technique is adopted to parallelize the OPT. The proposed approach is based on the Alternating Direction Method (ADM), a variant of the conventional Augmented Lagrangian approach, and makes it possible the independent regional AC-OPF for each control area while the global optimum for the entire system is assured. This paper is an extension of our previous work based on the auxiliary problem principle (APP). The proposed approach in this paper is a completely new one, however, in that ADM is based on the Proximal Point Algorithm which has long been recognized as one of the attractive methods for convex programming and min-max-convex-concave programming. The proposed method was demonstrated with IEEE 50-Bus system.