• Journal of Korea Society of Digital Industry and Information Management
• /
• v.11 no.4
• /
• pp.155-164
• /
• 2015

In this paper, we propose a new color interpolation method based on trilateral filter approach, which not only preserve the high-frequency components(image edge) while interpolating the missing raw data of color image(bayer data pattern), but also immune to the image noise components and better preserve the detail of the low-frequency components. The method is the trilateral filter approach applying a gradient to the low frequency components of the image signal in order to preserve the high-frequency components and the detail of the low-frequency components through the measure of the freedom of similarity among adjacent pixels. And also we perform Gaussian smoothing to the interpolated image data in order to robust to the noise. In this paper, we compare the conventional demosaicing algorithm and the proposed algorithm using 10 test images in terms of hue MAD, saturation MAD and CPSNR for the objective evaluation, and verify the performance of the proposed algorithm.

• ETRI Journal
• /
• v.24 no.3
• /
• pp.239-246
• /
• 2002

We investigate an alternative blind adaptive multiuser detection scheme based on a non-canonical linearly constrained constant modulus (LCCM) criterion and prove that, under the constrained condition, the non-canonical linearly constrained constant modulus algorithm (LCCMA) can completely remove multiple -access interference. We further demonstrate that the non-canonical LCCM criterion function is strictly convex in the noise-free state, and that under the constrained condition, it is also strictly convex even where small noise is present. We present a simple method for selecting the constant as well as a stochastic gradient algorithm for implementing our scheme. Numerical simulation results verify the scheme's efficiency.

• Journal of the Korea Society of Computer and Information
• /
• v.23 no.10
• /
• pp.23-31
• /
• 2018

Recently, there have been many studies on machine learning. Among them, studies on reinforcement learning are actively worked. In this study, we propose a controller to control bicycle using DDPG (Deep Deterministic Policy Gradient) algorithm which is the latest deep reinforcement learning method. In this paper, we redefine the compensation function of bicycle dynamics and neural network to learn agents. When using the proposed method for data learning and control, it is possible to perform the function of not allowing the bicycle to fall over and reach the further given destination unlike the existing method. For the performance evaluation, we have experimented that the proposed algorithm works in various environments such as fixed speed, random, target point, and not determined. Finally, as a result, it is confirmed that the proposed algorithm shows better performance than the conventional neural network algorithms NAF and PPO.

• Proceedings of the IEEK Conference
• /
• 2001.06e
• /
• pp.203-206
• /
• 2001

The fMRI which can express the function of brain as MR image is now being studied. The study on the functional image has usually been performed with the MRI in 4 tesla class in goneral, but if gradient echo imaging method could be used, it might make the most of what it has with the MRI in 1.5 tesla class. However, the lack of adequate image post-processing software prevents it from being used as widely as it could be. For the image post-processing algorithm of the functional image, subtraction method and several statistical methods are used with continuous introduction of new method recently. In this paper, we suggest adaptation of wavelet algorithm for obtaining a more reliable brain function map.

• The Transactions of the Korean Institute of Electrical Engineers
• /
• v.36 no.2
• /
• pp.73-79
• /
• 1987

This paper describes a new algorithm to evaluate the production cost for a generation system including energy-limited hydroelectric plants. The algorithm is based upon the analytical production costing model developed under the assumption of Gaussian probabilistic distribution of random load fluctuations and plant outages. Hydro operation and pumped storage operation have been dealt with in the previous papers using the concept of peak-shaving operation. In this paper, the hydro problem is solved by using a new version of the gradient projection method that treats the upper / lower bounds of variables saparately and uses a specified initial active constraint set. Accuracy and validity of the algorithm are demonstrated by comparing the result with that of the peak-shaving model.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2000.12a
• /
• pp.201-204
• /
• 2000

This paper describes a study on the time-varying control of parameters in learning of the neural network. Elman recurrent neural network (RNN) is used to implement the control of parameters. The parameters of learning and momentum rates In the error backpropagation algorithm ate updated at every iteration using fuzzy rules based on performance index. In addition, the gain and slope of the neuron's activation function are also considered time-varying parameters. These function parameters are updated using the gradient descent algorithm. Simulation results show that the auto-tuned learning algorithm results in faster convergence and lower system error than regular backpropagation in the system identification.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.16 no.12
• /
• pp.1156-1166
• /
• 2004

The shape of plate-fin type heat sink is numerically optimized to acquire the minimum pressure drop under the required temperature rise. In constrained nonlinear optimization problems of thermal/fluid systems, three fundamental difficulties such as high computational cost for function evaluations (i.e., pressure drop and thermal resistance), the absence of design sensitivity information, and the occurrence of numerical noise are commonly confronted. Thus, a sequential approximate optimization (SAO) algorithm has been introduced because it is very hard to obtain the optimal solutions of fluid/thermal systems by means of gradient-based optimization techniques. In this study, the progressive quadratic response surface method (PQRSM) based on the trust region algorithm, which is one of sequential approximate optimization algorithms, is used for optimization and the heat sink is optimized by combining it with the computational fluid dynamics (CFD).

• ETRI Journal
• /
• v.46 no.2
• /
• pp.205-217
• /
• 2024

We propose a bandwidth prediction approach based on deep learning. The approach is intended to accurately predict the bandwidth of various types of mobile networks. We first use a machine learning technique, namely, the gradient boosting algorithm, to recognize the connected mobile network. Second, we apply a handover detection algorithm based on network recognition to account for vertical handover that causes the bandwidth variance. Third, as the communication performance offered by 3G, 4G, and 5G networks varies, we suggest a bidirectional long short-term memory model with time distribution for bandwidth prediction per network. To increase the prediction accuracy, pretraining and fine-tuning are applied for each type of network. We use a dataset collected at University College Cork for network recognition, handover detection, and bandwidth prediction. The performance evaluation indicates that the handover detection algorithm achieves 88.5% accuracy, and the bandwidth prediction model achieves a high accuracy, with a root-mean-square error of only 2.12%.

• The Journal of the Acoustical Society of Korea
• /
• v.28 no.7
• /
• pp.640-645
• /
• 2009

CMA (Constant Modulus Algorithm) is one of the famous algorithms in blind channel equalization. Generally, CMA converges slowly and the speed of convergence is dependent on a step-size in the CMA procedure. Many researches have tried to speed up the convergence speed by applying a variable step-size to CMA. In this paper, we propose a new CMA algorithm with improved convergence performance. The improvement comes from an orthogonal projection of an average error gradient. We show the improvement in simulation results.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.19 no.3
• /
• pp.62-74
• /
• 2020

Although the government and local governments devote efforts to activate bicycles, they only access to the supply infrastructure such as bike lanes and the public bicycle rental service centers without considering the measures to overcome the geographical constraints of slope. Therefore, this study constructs bicyclist's energy consumption estimation model through experimental methods of slope gradient and heart rate measurement and suggest the bicycle route choice model which could minimize the energy by the slope gradient. After calculating the RMSE of the estimated energy consumption by applying this model to the simulation section, it is confirmed to be 41% better than the model which does not reflect slope gradient. The results of this study are expected to be applied to the bicycle infrastructure planning that considers both longitude and transverse of bike lanes and the algorithm of bicycle route guidance system in the future.