• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.42 no.3 s.303
• /
• pp.47-52
• /
• 2005

Obstructive Sleep Apnea (OSA) is a representative symptom of sleep disorder caused by the obstruction of upper airway. Because OSA causes not only excessive daytime sleepiness and fatigue, hypertension and arrhythmia but also cardiac arrest and sudden death during sleep in the severe case, it is very important to detect the occurrence and the frequency of OSA. OSA is usually diagnosed through the laboratory-based Polysomnography (PSG) which is uncomfortable and expensive. Therefore researches to improve the disadvantages of PSG are needed and studies for the detection of OSA using only one or two parameters are being made as alternatives to PSG. In this paper, we developed an algorithm for the detection of OSA based on Heart Rate Variability (HRV). The proposed method is applied to the ECG data sets provided from PhysioNet which consist of learning set and training set. We extracted features for the detection of OSA such as average and standard deviation of 1 minute R-R interval, power spectrum of R-R interval and S-peak amplitude from data sets. These features are applied to the input of neural network. As a result, we obtained sensitivity of $89.66\%$ and specificity of $95.25\%$. It shows that the features suggested in this study are useful to detect OSA.

• Journal of Korea Water Resources Association
• /
• v.51 no.11
• /
• pp.1021-1029
• /
• 2018

A primary objective of this study is to develop a drought forecasting technique based on groundwater which can be exploit for water supply under drought stress. For this purpose, we explored the lagged relationships between regionalized SGI (standardized groundwater level index) and SPI (standardized precipitation index) in view of the drought propagation. A regional prediction model was constructed using a NARX (nonlinear autoregressive exogenous) artificial neural network model which can effectively capture nonlinear relationships with the lagged independent variable. During the training phase, model performance in terms of correlation coefficient was found to be satisfactory with the correlation coefficient over 0.7. Moreover, the model performance was described by root mean squared error (RMSE). It can be concluded that the proposed approach is able to provide a reliable SGI forecasts along with rainfall forecasts provided by the Korea Meteorological Administration.

• Economic and Environmental Geology
• /
• v.53 no.3
• /
• pp.245-258
• /
• 2020

In this study, a data-driven response surface method using the results acquired from the numerical simulation is developed to evaluate the potential storage capacity of groundwater due to the construction of a groundwater dam. The hydraulic conductivities of alluvium and basement rock, depth and slope of the channel are considered as the natural conditions of the location for groundwater dam construction. In particular, the probability models of the hydraulic conductivities and the various types of geometry of the channel are considered to ensure the reliability of the numerical simulation and the generality of the developed estimation model. As the results of multiple simulations, it can be seen that the hydraulic conductivity of basement rock and the depth of the channel greatly influence to the groundwater storage capacity. In contrast, the slope of the channel along the groundwater flow direction shows a relatively lower impact on the storage capacity. Based on the considered natural conditions and the corresponding numerical simulation results, the storage capacity estimation model is developed applying an artificial neural network as the nonlinear regression model for training. The developed estimation model shows a high correlation coefficient (>0.9) between the simulated and the estimated storage amount. This result indicates the superiority of the developed model in evaluating the storage capacity of the potential location for groundwater dam construction without the numerical simulation. Therefore, a more objective and efficient comparison for the storage capacity between the different potential locations can be possibly made based on the developed estimation model. In line with this, the proposed method can be an effective tool to assess the optimal location of groundwater dam construction across Korea.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.19 no.2
• /
• pp.89-103
• /
• 2020

As we face an aging society, the demand for personal mobility for disabled and aged people is increasing. In fact, as of 2017, the number of electric wheelchair in the country continues to increase to 90,000. However, people with disabilities and seniors are more likely to have accidents while driving, because their judgment and coordination are inferior to normal people. One of the causes of the accident is the interference of personal vehicle steering control due to unbalanced road surface conditions. In this paper, we introduce a encoder type semantic segmentation algorithm that can recognize road conditions at high speed to prevent such accidents. To this end, more than 1,500 training data and 150 test data including road surface damage were newly secured. With the data, we proposed a deep neural network composed of encoder stages, unlike the Auto-encoding type consisting of encoder and decoder stages. Compared to the conventional method, this deep neural network has a 4.45% increase in mean accuracy, a 59.2% decrease in parameters, and an 11.9% increase in computation speed. It is expected that safe personal transportation will be come soon by utilizing such high speed algorithm.

• The Korean Journal of Applied Statistics
• /
• v.30 no.1
• /
• pp.41-55
• /
• 2017

This study predicts the average scores of top 150 PGA golf players on 132 PGA Tour tournaments (2013-2015) using data mining techniques and statistical analysis. This study also aims to predict the Top 10 and Top 25 best players in 4 different playoffs. Linear and nonlinear regression methods were used to predict average scores. Stepwise regression, all best subset, LASSO, ridge regression and principal component regression were used for the linear regression method. Tree, bagging, gradient boosting, neural network, random forests and KNN were used for nonlinear regression method. We found that the average score increases as fairway firmness or green height or average maximum wind speed increases. We also found that the average score decreases as the number of one-putts or scrambling variable or longest driving distance increases. All 11 different models have low prediction error when predicting the average scores of PGA Tournaments in 2015 which is not included in the training set. However, the performances of Bagging and Random Forest models are the best among all models and these two models have the highest prediction accuracy when predicting the Top 10 and Top 25 best players in 4 different playoffs.

• Journal of the Korea Convergence Society
• /
• v.12 no.9
• /
• pp.11-19
• /
• 2021

A software system is required to change during its life cycle due to various requirements such as adding functionalities, fixing bugs, and adjusting to new computing environments. Such program code modification should be considered as carefully as a new system development becase unexpected software errors could be introduced. In addition, when reusing open source programs, we can expect higher quality software if code changes of the open source program are predicted in advance. This paper proposes a Convolutional Neural Network (CNN)-based deep learning model to predict source code changes. In this paper, the prediction of code changes is considered as a kind of a binary classification problem in deep learning and labeled datasets are used for supervised learning. Java projects and code change logs are collected from GitHub for training and testing datasets. Software metrics are computed from the collected Java source code and they are used as input data for the proposed model to detect code changes. The performance of the proposed model has been measured by using evaluation metrics such as precision, recall, F1-score, and accuracy. The experimental results show the proposed CNN model has achieved 95% in terms of F1-Score and outperformed the multilayer percept-based DNN model whose F1-Score is 92%.

• Journal of the Korea Convergence Society
• /
• v.12 no.10
• /
• pp.63-70
• /
• 2021

Generating a super-resolution meteological data by using a high-resolution deep neural network can provide precise research and useful real-life services. We propose a new technique of generating improved training data for super-resolution deep neural networks. To generate high-resolution meteorological data with domain specific knowledge, Lambert conformal conic projection and objective analysis were applied based on observation data and ERA5 reanalysis field data of specialized institutions. As a result, temperature and humidity analysis data based on domain specific knowledge showed improved RMSE by up to 42% and 46%, respectively. Next, a super-resolution generative adversarial network (SRGAN) which is one of the aritifial intelligence techniques was used to automate the manual data generation technique using damain specific techniques as described above. Experiments were conducted to generate high-resolution data with 1 km resolution from global model data with 10 km resolution. Finally, the results generated with SRGAN have a higher resoltuion than the global model input data, and showed a similar analysis pattern to the manually generated high-resolution analysis data, but also showed a smooth boundary.

• Journal of the Korean Institute of Gas
• /
• v.22 no.6
• /
• pp.90-103
• /
• 2018

This paper presents the development of two ANN models which can predict an optimum production rate by controlling choke size in oil well, and gas injection rate in gas-lift well. The input data was solution gas-oil ratio, water cut, reservoir pressure, and choke size or gas injection rate. The output data was wellhead pressure and production rate. Firstly, a range of each parameters was decided by conducting sensitive analysis of input data for onshore oil well. In addition, 1,715 sets training data for choke size decision model and 1,225 sets for gas injection rate decision model were generated by nodal analysis. From the results of comparing between the nodal analysis and the ANN on the same reservoir system showed that the correlation factors were very high(>0.99). Mean absolute error of wellhead pressure and oil production rate was 0.55%, 1.05% with the choke size model, respectively. And the gas injection rate model showed the errors of 1.23%, 2.67%. It was found that the developed models had been highly accurate.

• Phonetics and Speech Sciences
• /
• v.11 no.1
• /
• pp.41-49
• /
• 2019

An end-to-end speech recognition model consisting of a single integrated neural network model was recently proposed. The end-to-end model does not need several training steps, and its structure is easy to understand. However, it is difficult to understand how the model recognizes speech internally. In this paper, we visualized and analyzed the attention-based end-to-end model to elucidate its internal mechanisms. We compared the acoustic model of the BLSTM-HMM hybrid model with the encoder of the end-to-end model, and visualized them using t-SNE to examine the difference between neural network layers. As a result, we were able to delineate the difference between the acoustic model and the end-to-end model encoder. Additionally, we analyzed the decoder of the end-to-end model from a language model perspective. Finally, we found that improving end-to-end model decoder is necessary to yield higher performance.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.24 no.4
• /
• pp.341-353
• /
• 2022

Pipelines are buried in urban area, and the position (depth and orientation) of buried pipeline should be clearly identified before ground excavation. Although various geophysical methods can be used to detect the buried pipeline, it is not easy to identify the exact information of pipeline due to heterogeneous ground condition. Among various non-destructive geo-exploration methods, ground penetration radar (GPR) can explore the ground subsurface rapidly with relatively low cost compared to other exploration methods. However, the exploration data obtained from GPR requires considerable experiences because interpretation is not intuitive. Recently, researches on automated detection technology for GPR data using deep learning have been conducted. However, the lack of GPR data which is essential for training makes it difficult to build up the reliable detection model. To overcome this problem, we conducted a preliminary study to improve the performance of the detection model using finite difference time domain (FDTD)-based numerical analysis. Firstly, numerical analysis was performed with homogeneous soil media having single permittivity. In case of heterogeneous ground, numerical analysis was performed considering the ground heterogeneity using fractal technique. Secondly, deep learning was carried out using convolutional neural network. Detection Model-A is trained with data set obtained from homogeneous ground. And, detection Model-B is trained with data set obtained from homogeneous ground and heterogeneous ground. As a result, it is found that the detection Model-B which is trained including heterogeneous ground shows better performance than detection Model-A. It indicates the ground heterogeneity should be considered to increase the performance of automated detection model for GPR exploration.