• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.25 no.5
• /
• pp.285-290
• /
• 2013

Both full linear and shallow water edge waves are compared to get a better understanding of edge wave behavior. By using method of separation of variables, we are able to get solution of full linear edge wave presented by Ursell (1952) without derivation. The shallow water edge waves show dispersive features despite being derived from shallow water equations. When bottom slope is mild enough, shallow water edge wave tends to linear edge wave and has some advantages of manipulation. Solution of edge wave generated by a moving landslide of Gaussian shape is constructed by an expansion of shallow water normal modes. Numerical results are presented and discussed on their main features.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.22 no.6
• /
• pp.793-798
• /
• 2012

In this paper, we explored the new method for extracting feature from the electroencephalography (EEG) signal based on linear regression technique with the orthonormal polynomial bases. At first, EEG signals from electrodes around motor cortex were selected and were filtered in both spatial and temporal filter using band pass filter for alpha and beta rhymic band which considered related to the synchronization and desynchonization of firing neurons population during motor imagery task. Signal from epoch length 1s were fitted into linear regression with Legendre polynomials bases and extract the linear regression weight as final features. We compared our feature to the state of art feature, power band feature in binary classification using support vector machine (SVM) with 5-fold cross validations for comparing the classification accuracy. The result showed that our proposed method improved the classification accuracy 5.44% in average of all subject over power band features in individual subject study and 84.5% of classification accuracy with forward feature selection improvement.

• Journal of information and communication convergence engineering
• /
• v.22 no.2
• /
• pp.165-171
• /
• 2024

In this study, we present a novel approach for enhancing chest X-ray image classification (normal, Covid-19, edema, mass nodules, and pneumothorax) by combining contrastive learning and machine learning algorithms. A vast amount of unlabeled data was leveraged to learn representations so that data efficiency is improved as a means of addressing the limited availability of labeled data in X-ray images. Our approach involves training classification algorithms using the extracted features from a linear fine-tuned Momentum Contrast (MoCo) model. The MoCo architecture with a Resnet34, Resnet50, or Resnet101 backbone is trained to learn features from unlabeled data. Instead of only fine-tuning the linear classifier layer on the MoCopretrained model, we propose training nonlinear classifiers as substitutes for softmax in deep networks. The empirical results show that while the linear fine-tuned ImageNet-pretrained models achieved the highest accuracy of only 82.9% and the linear fine-tuned MoCo-pretrained models an increased highest accuracy of 84.8%, our proposed method offered a significant improvement and achieved the highest accuracy of 87.9%.

• Proceedings of the KGS Conference
• /
• 2005.05a
• /
• pp.37-39
• /
• 2005

• Journal of the Korean Society for Nondestructive Testing
• /
• v.29 no.6
• /
• pp.563-569
• /
• 2009

In this paper, four different types of pulsed eddy current(PEC) probe are designed and their performance of detecting wall thickness reduction is compared. By using the backward difference method in time and the finite element method in space, PEC signals from various thickness and materials are numerically calculated and three features of the signal are selected. Since PEC signals and features are obtained by various types and sizes of probe, the comparison is made through the normalized features which reflect the sensitivity of the feature to thickness reduction. The normalized features indicate that the shielded reflection probe provides the best sensitivity to wall thickness reduction for all three signal features. Results show that the best sensitivity to thickness reduction can be achieved by the peak value, but also suggest that the time to peak can be a good candidate because of its linear relationship with the thickness variation.

• Journal of Biosystems Engineering
• /
• v.19 no.4
• /
• pp.380-396
• /
• 1994

A personal computer based color machine vision system with video camera and fluorescent lighting system was used to generate images of stationary tobacco leaves. Image processing algorithms were developed to extract both the geometric and the color features of tobacco leaves. Geometric features include area, perimeter, centroid, roundness and complex ratio. Color calibration scheme was developed to convert measured pixel values to the standard color unit using both statistics and artificial neural network algorithm. Improved back propagation algorithm showed less sum of square errors than multiple linear regression. Color features provide not only quality evaluation quantities but the accurate color measurement. Those quality features would be useful in grading tobacco automatically. This system would also be useful in measuring visual features of other agricultural products.

• Journal of the Institute of Convergence Signal Processing
• /
• v.13 no.3
• /
• pp.136-141
• /
• 2012

One of the main reasons for serious road accidents is driving while drowsy. For this reason, drowsiness detection and warning system for drivers has recently become a very important issue. Monitoring physiological signals provides the possibility of detecting features of drowsiness and fatigue of drivers. One of the effective signals is to measure electroencephalogram (EEG) signals and electrooculogram (EOG) signals. The aim of this study is to extract drowsiness-related features from a set of EEG signals and to classify the features into three states: alertness, drowsiness, sleepiness. This paper proposes a neural-network-based drowsiness detection system using Linear Predictive Coding (LPC) coefficients as feature vectors and Multi-Layer Perceptron (MLP) as a classifier. Samples of EEG data from each predefined state were used to train the MLP program by using the proposed feature extraction algorithms. The trained MLP program was tested on unclassified EEG data and subsequently reviewed according to manual classification. The classification rate of the proposed system is over 96.5% for only very small number of samples (250ms, 64 samples). Therefore, it can be applied to real driving incident situation that can occur for a split second.

• Journal of applied mathematics & informatics
• /
• v.10 no.1_2
• /
• pp.17-26
• /
• 2002

A stability analysis of a non-linear prey-predator system under the influence of one dimensional diffusion has been investigated to determine the nature of the bifurcation point of the system. The non-linear bifurcation analysis determining the steady state solution beyond the critical point enables us to determine characteristic features of the spatial inhomogeneous pattern arising out of the bifurcation of the state of the system.

• Journal of Mechanical Science and Technology
• /
• v.20 no.2
• /
• pp.205-211
• /
• 2006

Among other critical conditions in rotor systems the large non-linear vibration excited by bearing non-linearity causes the rotor failure. For reducing this catastrophic failure and predictive detection of this phenomenon the analysis of orthotropic bearing non-linearity in rotor system using higher order frequency response functions (HFRFs) is conducted and is shown to be theoretically feasible as that of non-rotating structures. The complex HFRFs based on the Volterra series are newly developed for the process and investigated their features by using the simple forms of the FRFs associated with the forward and the backward modes.

• Korean Journal of Geomatics
• /
• v.5 no.1
• /
• pp.7-19
• /
• 2005

Photogrammetric mapping procedures have gone through major developments due to significant improvements in its underlying technologies. The availability of GPS/INS systems greatly assist in direct geo-referencing of the acquired imagery. Still, photogrammetric datasets taken without the aid of positioning and navigation systems need control information for the purpose of surface reconstruction. Point features were, and still are, the primary source of control for the photogrammetric triangulation although other higher-order features are available and can be used. LIDAR systems supply dense geometric surface information in the form of three dimensional coordinates with respect to certain reference system. Considering the accuracy improvement of LIDAR systems in the recent years, LIDAR data is considered a viable supply of photogrammetric control. To exploit LIDAR data, new challenges are poised concerning the representation and reference system by which both the photogrammetric and LIDAR datasets are described. In this paper, registration methodologies will be devised for the purpose of integrating the LIDAR data into the photogrammetric triangulation. Such registration methodologies have to deal with three issues: registration primitives, transformation parameters, and similarity measures. Two methodologies will be introduced that utilize straight-line and areal features derived from both datasets as the registration primitives. The first methodology directly incorporates the LIDAR lines as control information in the photogrammetric triangulation, while in the second methodology, LIDAR patches are used to produce and align the photogrammetric model. Also, camera self-calibration experiments were conducted on simulated and real data to test the feasibility of using LIDAR patches for this purpose.