• Smart Media Journal
• /
• v.9 no.2
• /
• pp.22-32
• /
• 2020

The hippocampal volume atrophy is known to be linked with neuro-degenerative disorders and it is also one of the most important early biomarkers for Alzheimer's disease detection. The measurements of hippocampal pure volumes from Magnetic Resonance Imaging (MRI) is a crucial task and state-of-the-art methods require a large amount of time. In addition, the structural brain development is investigated using MRI data, where brain morphometry (e.g. cortical thickness, volume, surface area etc.) study is one of the significant parts of the analysis. In this study, we have proposed a patch-based ensemble model of 3-D convolutional neural network (CNN) to measure the hippocampal pure volume from MRI data. The 3-D patches were extracted from the volumetric MRI scans to train the proposed 3-D CNN models. The trained models are used to construct the ensemble 3-D CNN model and the aggregated model predicts the pure volume in one-step in the test phase. Our approach takes only 5 seconds to estimate the volumes from an MRI scan. The average errors for the proposed ensemble 3-D CNN model are 11.7±8.8 (error%±STD) and 12.5±12.8 (error%±STD) for the left and right hippocampi of 65 test MRI scans, respectively. The quantitative study on the predicted volumes over the ground truth volumes shows that the proposed approach can be used as a proxy.

• Journal of Korean Society on Water Environment
• /
• v.24 no.1
• /
• pp.91-98
• /
• 2008

This work was performed to develop a model possible to predict the influent flow and influent components, which are one of main disturbances causing process problems at the operation of municipal wastewater treatment plant. In this study, artificial neural network (ANN) was used in order to develop a model that was able to predict the influent flow, $COD_{Mn}$, SS, TN 1 day-ahead, 2day-ahead and 3 day ahead. Multi-layer feed-forward back-propagation network was chosen as neural network type, and tanh-sigmoid function was used as activation function to transport signal at the neural network. And Levenberg-Marquart (LM) algorithm was used as learning algorithm to train neural network. Among 420 data sets except missing data, which were collected between 2005 and 2006 at field plant, 210 data sets were used for training, and other 210 data sets were used for validation. As result of it, ANN model for predicting the influent flow and components 1-3day ahead could be developed successfully. It is expected that this developed model can be practically used as follows: Detecting the fault related to effluent concentration that can be happened in the future by combining with other models to predict process performance in advance, and minimization of the process fault through the establishment of various control strategies based on the detection result.

• Proceedings of the KIEE Conference
• /
• 1998.07c
• /
• pp.1166-1169
• /
• 1998

The Load characteristic of electric railway requires the power demand of the high capacity which amplitude is spacial-temporally fluctuated due to frequent starting and stopping with large tractive force. The conventional electric railway mainly consists of the resistance controlled and the thyristor controlled locomotives, are compensated for their bad characteristics of the power factor$(70\sim80%)$ with installation of another capacitor improving power factor at the substation. Since 1994, VVVF train car with good characteristics of power factor(100%) have been introduced and operated in Kwa-Chon Line. From the present technical tendency, it is judged that introduction of the locomotive with various controlled methods is necessary. The protective equipments installed at the substation are complicated and various aspects to detect faults and reduce their extension, so the universal countermeasures are required. Specially in the case of the fault occurrence it is difficult to calculate the fault location because of the change in the contactline constant according to modifying the characteristics of the contactline (the dualized catenary wire and extension, etc), so much time is required for the detection of fault location. In BT-fed method distance-relays and fault-locators are not installed, we have so many difficulties in the quick accident recovery.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.3
• /
• pp.312-318
• /
• 2012

Recently, although the need of marine robots being raised in extreme areas, the basis is very deficient. Fortunately, as the robot competition is vitalizing and the need of the robot education is increasing, it is desirable to establish the basis of the R&D and industrialization of marine robots and to train professionals through the development and diffusion of marine robot kits. However, in conventional case, there is no underwater-type autonomous marine robot kit for the marine robot competition, which has the abilities of the underwater locomotion and target detection and avoidance. To solve this problem, a marine robot kit which has the abilities of the underwater locomotion, the waterproof and the weight adjustment, is developed. To verify the performance of the developed kit, test and evaluation such as surge, pitch, yaw, obstacle avoidance is performed. The test and evaluation results show that the possibility of the real applications of the developed kit.

• Journal of the Institute of Convergence Signal Processing
• /
• v.10 no.4
• /
• pp.233-238
• /
• 2009

In this paper, we propose the methods to select the stable region for the detect region which is required in the system used the face to the ROI in the auto-focus digital camera. this method regards the face region as the ROI in the progressive input frame and focusing the region in the mobile camera embeded ISP module automatically. The learning algorithm to detect the face is the Adaboost algorithm. we proposed the method to detect the slanted face not participate in the train process and postprocessing method for the results of detection, and then we proposed the stabilization method to sustain the region not shake for the region. we estimated the capability for the stabilization algorithm using the RMS between the trajectory and regression curve.

• Fashion & Textile Research Journal
• /
• v.21 no.1
• /
• pp.88-95
• /
• 2019

The purpose of this study was to figure out the various characteristics of the thickness sensation among elements of tactile sensation using psychophysical method. Firstly, panel screening was processed to select sensitive thickness panel using the triangle test. As a result of discriminating the paper thickness difference from 1 to 4 pieces, the female students perceived the thickness difference more sensitively than the male students (p<.05). Secondly, JND (Just Noticeable Difference) was obtained at percentage of stimulus detection rate in order to detect the degree of thickness difference by psychophysical method. It was found that the difference threshold of the entire group was about 0.125mm, with male group being about 0.178mm and female group being about 0.095mm. Thirdly, Weber's law was used to find the minimum discrimination difference between the stimuli. The experiments were conducted by increasing the paper's base thickness from 1.950mm to 1.330mm, and it was found that the difference tendency increased when the size of the basic stimulus increased. At this time, the minimum discrimination difference increased, but the Weber' fraction was not proportional to the magnitude of the stimulus. The significance of this study was that sensory evaluation was applied to research in the fields of clothing science and it seems effective to further screen and train sensitive students as material discrimination experts.

• Journal of the Korea Convergence Society
• /
• v.12 no.1
• /
• pp.49-55
• /
• 2021

Recently, with the development of database, it is possible to store a lot of data generated in finance, security, and networks. These data are being analyzed through classifiers based on machine learning. The main problem at this time is data imbalance. When we train imbalanced data, it may happen that classification accuracy is degraded due to over-fitting with majority class data. To overcome the problem of data imbalance, oversampling strategy that increases the quantity of data of minority class data is widely used. It requires to tuning process about suitable method and parameters for data distribution. To improve the process, In this study, we propose a strategy to explore and optimize oversampling combinations and ratio based on various methods such as synthetic minority oversampling technique and generative adversarial networks through genetic algorithms. After sampling credit card fraud detection which is a representative case of data imbalance, with the proposed strategy and single oversampling strategies, we compare the performance of trained classifiers with each data. As a result, a strategy that is optimized by exploring for ratio of each method with genetic algorithms was superior to previous strategies.

• Journal of Convergence for Information Technology
• /
• v.11 no.7
• /
• pp.14-20
• /
• 2021

Most accidents caused by road icing in winter lead to major accidents. Because it is difficult for the driver to detect the road icing in advance. In this work, we study how to accurately detect road traffic emerging risk using AutoML and CNN's ensemble model that use both structured and unstructured data. We train CNN-based road traffic emerging risk classification model using images that are unstructured data and AutoML-based road traffic emerging risk classification model using weather data that is structured data, respectively. After that the ensemble model is designed to complement the CNN-based classification model by inputting probability values derived from of each models. Through this, improves road traffic emerging risk classification performance and alerts drivers more accurately and quickly to enable safe driving.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.1
• /
• pp.111-118
• /
• 2022

In computer vision research, the two-dimensional human pose is a very extensive research direction, especially in pose tracking and behavior recognition, which has very important research significance. The acquisition of human pose targets, which is essentially the study of how to accurately identify human targets from pictures, is of great research significance and has been a hot research topic of great interest in recent years. Human pose recognition is used in artificial intelligence on the one hand and in daily life on the other. The excellent effect of pose recognition is mainly determined by the success rate and the accuracy of the recognition process, so it reflects the importance of human pose recognition in terms of recognition rate. In this human body gesture recognition, the human body is divided into 17 key points for labeling. Not only that but also the key points are segmented to ensure the accuracy of the labeling information. In the recognition design, use the comprehensive data set MS COCO for deep learning to design a neural network model to train a large number of samples, from simple step-by-step to efficient training, so that a good accuracy rate can be obtained.

• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.341-348
• /
• 2022

Wi-Fi Sensing, which uses Wi-Fi technology to sense the surrounding environments, has strong potentials in a variety of sensing applications. Recently several advanced deep learning-based solutions using CSI (Channel State Information) data have achieved high performance, but it is still difficult to use in practice without explicit data collection, which requires expensive adaptation efforts for model retraining. In this study, we propose a Channel State Information Automatic Labeling System (CALS) that automatically collects and labels training CSI data for deep learning-based Wi-Fi sensing systems. The proposed system allows the CSI data collection process to efficiently collect labeled CSI for labeling for supervised learning using computer vision technologies such as object detection algorithms. We built a prototype of CALS to demonstrate its efficiency and collected data to train deep learning models for detecting the presence of a person in an indoor environment, showing to achieve an accuracy of over 90% with the auto-labeled data sets generated by CALS.