• The Journal of the Korea Contents Association
• /
• v.12 no.2
• /
• pp.1-9
• /
• 2012

This paper presents an algorithm for depression diagnosis using the Neural Network with Weighted Fuzzy Membership functions (NEWFM) and heart rate variability (HRV). In the algorithm, 22 different features were initially extracted from the HRV signal by frequency domain, time domain, wavelet transformed, and Poincar$\acute{e}$ transformed feature extraction methods; of these 6 optimal features were selected by significance evaluation using Non-overlap Area Distribution Measurement (NADM) based on NEWFM. The proposed algorithm uses these 6 optimal features to diagnose depression with an accuracy of 95.83%.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.153-166
• /
• 2022

Identifying fine cracks in steel bridge facilities is a challenging task of structural health monitoring (SHM). This study proposed an end-to-end crack image segmentation framework based on a one-step Convolutional Neural Network (CNN) for pixel-level object recognition with high accuracy. To particularly address the challenges arising from small object detection in complex background, efforts were made in loss function selection aiming at sample imbalance and module modification in order to improve the generalization ability on complicated images. Specifically, loss functions were compared among alternatives including the Binary Cross Entropy (BCE), Focal, Tversky and Dice loss, with the last three specialized for biased sample distribution. Structural modifications with dilated convolution, Spatial Pyramid Pooling (SPP) and Feature Pyramid Network (FPN) were also performed to form a new backbone termed CrackDet. Models of various loss functions and feature extraction modules were trained on crack images and tested on full-scale images collected on steel box girders. The CNN model incorporated the classic U-Net as its backbone, and Dice loss as its loss function achieved the highest mean Intersection-over-Union (mIoU) of 0.7571 on full-scale pictures. In contrast, the best performance on cropped crack images was achieved by integrating CrackDet with Dice loss at a mIoU of 0.7670.

• Proceedings of the IEEK Conference
• /
• 2000.11a
• /
• pp.469-472
• /
• 2000

Security-related issues in mobile communications are increasing. The security requirements of mobile communications for the mobile users include authentication of the mobile user, the data confidentiality, the data confidentiality and the location privacy of mobile user. These services require security features compatible with the wireline networks. However, wireless networks have many restrictions compare to wireline networks such as the limited computational capability of mobile equipment and limited resource(bandwidth) between a mobile user and a fixed network. So, security features for IMT-2000 are designed to meet the limited capacity. In this paper, we analyze the required security features and mechanism, and design network access security feature effective for IMT-2000 Systems. The design includes security functions allocation to each system. Finally, discuss the computational power of each system based on at]coated functions to it

• Industrial Engineering and Management Systems
• /
• v.15 no.2
• /
• pp.148-155
• /
• 2016

Existing power plants may consume significant amounts of fuel and require high operating costs, partly because of poor electrical power output estimates. This paper suggests a continuous conditional random field (C-CRF) model to predict more precisely the full-load electrical power output of a base load operated combined cycle power plant. We introduce three feature functions to model association potential and one feature function to model interaction potential. Together, these functions compose the C-CRF model, and the model is transformed into a multivariate Gaussian distribution with which the operation parameters can be modeled more efficiently. The performance of our model in estimating power output was evaluated by means of a real dataset and our model outperformed existing methods. Moreover, our model can be used to estimate confidence intervals of the predicted output and calculate several probabilities.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1995.10a
• /
• pp.711-715
• /
• 1995

In the concept design stage of the product design process, it is desirable that a designer makes alternative designs sufficiently, examines and analyzes them, and finally determines an appropriate design. To efficiently investigate several alternative designs, it should be facilitated to modify the model and transfer the model data to analysis program. In this research, a concept design process for tank is automated using I-DEAS feature-based modeling system from SDRC. Additionally, the facility for the pre-estimation of the performance of product, the useful volume calculation, the mass calculation, the confirmation of the allowable workspae, and the interface to analysis propram are developed using API functions of OPen-link and Open-data. Graphic User Interface (GUI) makes it extrmely easy to utilize functions.

• Proceedings of the Korean Society for Technology of Plasticity Conference
• /
• 2002.02a
• /
• pp.3-10
• /
• 2002

In this study on machine measuring system based on solid feature was developed. This system was applied with injection mold using 3 dimensional solid modeler for verification. Developed program include pre-processor, main processor, and post processor. In pre-processor there are functions which check intersection, simulate motion of probe and calculate measuring time. Main processor generates measuring path and output NC code in Unigraphics. In post-processor functions that include evaluation of undercut or overcut and display of measuring procedure are offered. In addition analysis module for quality control of measured data on manufactured product was developed with geometric and dimensional tolerance concept. As the result developed program could get stability of system, precision of product, rapidity and cost down of manufacturing process compared with before measuring process.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.20 no.10
• /
• pp.900-906
• /
• 2010

This note proposes vibration-free motor control through modified LQG/LTR methodology. A conventional LQG/LTR method is a design tool in the frequency domain. However, unlike the conventional one, the proposed one is a time response based design method. This feature is firstly designed by parameterized settling time control gain through the target loop design procedure and the feature is secondly realized by loop transfer recovery. In order to show convergence to the target loop transfer functions, asymptotic behaviors of the open and the closed loop transfer functions are shown. At the conclusion, it is verified that the proposed method is robustly stable to parametric uncertainties through ${\mu}$-plot.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.9
• /
• pp.4372-4389
• /
• 2019

Spectrum sensing has attracted much attention due to its significant contribution to idle spectrum detection in Cognitive Radio Networks. However, specialized discussion is on complexity-based sensing strategy for spectrum sensing seldom considered. Motivated by this, this paper is devoted to complexity-based sensing strategy for spectrum sensing. Firstly, three efficiency functions are defined to estimate sensing efficiency of a spectrum scheme. Then a novel sensing strategy is proposed given sensing performance and computational complexity. After that, the proposed sensing strategy is extended to energy detector, Cyclostationary feature detector, covariance matrix detector and cooperative spectrum detector. The proposed sensing strategy provides a novel insight into sensing performance estimation for its consideration of both sensing capacity and sensing complexity. Simulations analyze three efficiency functions and optimal sensing strategy of energy detector, Cyclostationary feature detector and covariance matrix detector.

• Animal cells and systems
• /
• v.2 no.2
• /
• pp.153-164
• /
• 1998

Cytokines regulate proliferation, differentiation and functions of haemotopoietic cells. Each cytokine possesses a variety of activities on various target cells (pleiotropy) and various cytokines have similar and overlapping activities on the same target cells (redundancy). The nature of these cytokine activities predicts unique feature of cytokine receptors, namely, cytokine has multiple receptors, different cytokines share a common receptor, and different cytokine receptors are linked to common signaling pathways. cDNA cloning of genes for cytokine receptors revealed distinct sets of receptor family with different structural features. The cytokine receptor superfamily consists of a largest family, and contains more than twenty cytokine receptor subunits. This receptor has common structural features in both extracellular and intracellular regions without tyrosine kinase domain. Another striking feature of the receptor is to share common subunit of multiple cytokines, which partly explains the redundancy of activities of some cytokines. Recent studies revealed detailed signaling events of the cytokine receptor, the primary activation of JAK and subsequent phosphorylation of tyrosine residues of receptor, and various cellular proteins. Many SH2 containing adapter proteins play an important role in cytokine signals, and this system has similarities with tyrosine kinase receptor signal transduction. STAT may mainly account for cytokine specific functions as suggested by knockout mice studies. It is of importance to note that cytokine activates multiple signaling pathways and the balance and combination of related signaling events may determine the specificity of functions of cytokines.

• Journal of Information Processing Systems
• /
• v.2 no.1
• /
• pp.23-27
• /
• 2006

This paper presents a feature selection technique based on distributional differences for efficient machine learning. Initial training data consists of data including many features and a target value. We classified them into positive and negative data based on the target value. We then divided the range of the feature values into 10 intervals and calculated the distribution of the intervals in each positive and negative data. Then, we selected the features and the intervals of the features for which the distributional differences are over a certain threshold. Using the selected intervals and features, we could obtain the reduced training data. In the experiments, we will show that the reduced training data can reduce the training time of the neural network by about 40%, and we can obtain more profit on simulated stock trading using the trained functions as well.