• Journal of Korea Society of Industrial Information Systems
• /
• v.27 no.4
• /
• pp.19-27
• /
• 2022

In general, the performance of ML(Machine Learning) application is determined by various factors such as the type of ML model, the size of model (number of parameters), hyperparameters setting during the training, and training data. In particular, the recognition accuracy of ML may be deteriorated or experienced overfitting problem if the amount of dada used for training is insufficient. Existing studies focusing on image recognition have widely used open datasets for training and evaluating the proposed ML models. However, for specific applications where the sensor used, the target of recognition, and the recognition situation are different, it is necessary to build the dataset manually. In this case, the performance of ML largely depends on the quantity and quality of the data. In this paper, training data used for motion recognition application is augmented using the kernel density estimation algorithm which is a type of non-parametric estimation method. We then compare and analyze the recognition accuracy of a ML application by varying the number of original data, kernel types and augmentation rate used for data augmentation. Finally experimental results show that the recognition accuracy is improved by up to 14.31% when using the narrow bandwidth Tophat kernel.

• The Journal of the Korea Contents Association
• /
• v.14 no.4
• /
• pp.9-18
• /
• 2014

Recently, many researches have been focused on multi-view video applications and services such as wireless video surveillance networks, wireless video sensor networks and wireless mobile video. In multi-view video signal processing, to exploit the strong correlation between images acquired by different cameras plays great role in developing a core technique of multi-view video coding. This paper proposes an adaptive multi-view video interpolation technique which is applicable for multi-view distributed video coding without requiring any cooperation amongst the cameras. The proposed algorithm estimates the non-linear moving blocks and employs disparity compensated view prediction, and then fills in the unreliable blocks. Through computer simulations, it is shown that the proposed method outperforms the conventional methods.

• Journal of Korean Society on Water Environment
• /
• v.34 no.4
• /
• pp.391-398
• /
• 2018

The main purpose of this study is to suggest and recommend the more reliable flow direction methods within the framework of DEM and power law distribution, by investigating the existing methodologies. To this end SFD (single flow direction method), MFD (multiple flow direction method) and IFD (Infinite flow direction method) are applied to analyze the determination of a flow direction for the water particles as seen in the Jeonjeokbigyo basin, and then assessed with respect to the variation of flow accumulation in that region. As the main results revealed, the study showed the different patterns of flow accumulation are found out from each applications of flow direction methods utilized in this study. This brings us to understand that as the flow dispersion on DEM increases, in this case the contributing areas to the outlet grow in sequence of SFD, IFD, MFD, but it is noted that the contribution of individual pixels into outlet decreases at that time. In what follows, especially with the MFD and IFD, the result tends to make additional hydrologic abstraction from rainfall excess, as noted due to the flow dispersion within flow paths on DEM. Based on the parameter estimation for a power law distribution, which is frequently used for identify the aggregation structure of complex system, by maximum likelihood flow accumulation can be thought of as a scale invariance factor. In this regard, the combination of flow direction methods could give rise to the more realistic water flow on DEM, as revealed through the separate flow direction methods as utilized for dispersion and aggregation effects of water flow within the available different topographies.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.1
• /
• pp.177-185
• /
• 2009

Cluster analysis or clustering is a kind of unsupervised learning method in which a set of data points is divided into a given number of homogeneous groups. Fuzzy clustering method, one of the most popular clustering method, allows a point to belong to all the clusters with different degrees, so produces more intuitive and natural clusters than hard clustering method does. Even more some of fuzzy clustering variants have noise-immunity. In this paper, we improved the Possibilistic Fuzzy C-Means (PFCM), which generates a membership matrix as well as a typicality matrix, using Gath-Geva (GG) method. The proposed method has a focus on the boundaries of clusters, which is different from most of the other methods having a focus on the centers of clusters. The generated membership values are suitable for the classification-type applications. As the typicality values generated from the algorithm have a similar distribution with the values of density function of Gaussian distribution, it is useful for Gaussian-type density estimation. Even more GG method can handle the clusters having different numbers of data points, which the other well-known method by Gustafson and Kessel can not. All of these points are obvious in the experimental results.

• Progress in Superconductivity and Cryogenics
• /
• v.17 no.1
• /
• pp.10-13
• /
• 2015

High-temperature superconducting coated conductors (HTS-CCs) are good candidates for resistive superconducting fault current limiter (RSFCL) applications. However, the high current density they can carry and their low thermal diffusivity expose them to the risk of thermal instability. In order to find the best compromise between stability and cost, it is important to study the heat transfer between HTS-CCs and the liquid nitrogen ($LN_2$) bath. This paper presents an experimental method to monitor in real-time the temperature of a quenched HTS-CC during a current pulse. The current and the associated voltage are measured, giving a precise knowledge of the amount of energy dissipated in the tape. These values are compared with an adiabatic numerical thermal model which takes into account heat capacity temperature dependence of the stabilizer and substrate. The result is a precise estimation of the heat transfer to the liquid nitrogen bath at each time step. Measurements were taken on a bare tape and have been repeated using increasing $Kapton^{(R)}$ insulation layers. The different heat exchange regimes can be clearly identified. This experimental method enables us to characterize the recooling process after a quench. Finally, suggestions are done to reduce the temperature increase of the tape, at a rated current and given limitation time, using different thermal insulation thicknesses.

• Journal of Communications and Networks
• /
• v.16 no.3
• /
• pp.344-354
• /
• 2014

Transmission control protocol friendly rate control (TFRC) is designed to mainly provide optimal service for unicast applications, such as multimedia streaming in the best-effort Internet environment. However, high bandwidth networks with large delays present an environment where TFRC may have a problem in utilizing the full bandwidth. TFRC inherits the slow-start mechanism of TCP Reno, but this is a time-consuming process that may require many round-trip-times (RTTs), until an appropriate sending rate is reached. Another disadvantage inherited from TCP Reno is the RTT-unfairness problem, which severely affects the performance of long-RTT flows. In this paper, we suggest enhanced TFRC for high quality video streaming over high bandwidth delay product networks. First, we propose a fast startup scheme that increases the data rate more aggressively than the slow-start, while mitigating the overshooting problem. Second, we propose a bandwidth estimation method to achieve more equitable bandwidth allocations among streaming flows that compete for the same narrow link with different RTTs. Finally, we improve the responsiveness of TFRC in the presence of severe congestion. Simulation results have shown that our proposal can achieve a fast startup and provide fairness with competing flows compared to the original TFRC.

• Proceedings of the KSRS Conference
• /
• 2002.10a
• /
• pp.40-44
• /
• 2002

Precision correction is the process of geometrically aligning images to a reference coordinate system using GCPs(Ground Control Points). Many applications of remote sensing data, such as change detection, mapping and environmental monitoring, rely on the accuracy of precision correction. However it is a very time consuming and laborious process. It requires GCP collection, the identification of image points and their corresponding reference coordinates. At typical satellite ground stations, GCP collection requires most of man-powers in processing satellite images. A method of automatic registration of satellite images is demanding. In this paper, we propose a new algorithm for automatic precision correction by GCP chips and RANSAC(Random Sample Consensus). The algorithm is divided into two major steps. The first one is the automated generation of ground control points. An automated stereo matching based on normalized cross correlation will be used. We have improved the accuracy of stereo matching by determining the size and shape of match windows according to incidence angle and scene orientation from ancillary data. The second one is the robust estimation of mapping function from control points. We used the RANSAC algorithm for this step and effectively removed the outliers of matching results. We carried out experiments with SPOT images over three test sites which were taken at different time and look-angle with each other. Left image was used to select UP chipsets and right image to match against GCP chipsets and perform automatic registration. In result, we could show that our approach of automated matching and robust estimation worked well for automated registration.

• International Journal of Reliability and Applications
• /
• v.13 no.1
• /
• pp.19-35
• /
• 2012

This paper presents an optimum design of step-stress partially accelerated life test (PALT) plan which allows the test condition to be changed from use to accelerated condition on the occurrence of fixed number of failures. Various life distribution models such as exponential, Weibull, log-logistic, Burr type-Xii, etc have been used in the literature to analyze the PALT data. The need of different life distribution models is necessitated as in the presence of a limited source of data as typically occurs with modern devices having high reliability, the use of correct life distribution model helps in preventing the choice of unnecessary and expensive planned replacements. Truncated distributions arise when sample selection is not possible in some sub-region of sample space. In this paper it is assumed that the lifetimes of the items follow Truncated Logistic distribution truncated at point zero since time to failure of an item cannot be negative. Optimum step-stress PALT plan that finds the optimal proportion of units failed at normal use condition is determined by using the D-optimality criterion. The method developed has been explained using a numerical example. Sensitivity analysis and comparative study have also been carried out.

• Asia Marketing Journal
• /
• v.11 no.1
• /
• pp.29-38
• /
• 2009

Estimating the Bass diffusion model often creates a time-interval bias, which leads the OLS approach to overestimate sales at early stages and underestimate sales after the peak. Further, a specification error from omitted variables might raise serial correlations among residuals when marketing actions are not incorporated into the diffusion model. Autocorrelated disturbances may yield unbiased but inefficient estimation, and therefore invalid inference results. This phenomenon warrants a modified approach to estimating the Bass diffusion model. In this paper, the authors propose a modified Bass diffusion model handling autocorrelated disturbances. To validate the new approach, authors applied the method on two different data-sets: CT Scanners in the U.S, and FPD TV sales in Korea. The results showed improved model fit and the validity of the proposed model.

• Steel and Composite Structures
• /
• v.25 no.6
• /
• pp.663-670
• /
• 2017

The problem of layup optimization of the composite laminates involves a very complex multidimensional solution space which is usually non-exhaustively explored using different heuristic computational methods such as genetic algorithms (GA). To ensure the convergence to the global optimum of the applied heuristic during the optimization process it is necessary to evaluate a lot of layup configurations. As a consequence the analysis of an individual layup configuration should be fast enough to maintain the convergence time range to an acceptable level. On the other hand the mechanical behavior analysis of composite laminates for any geometry and boundary condition is very convoluted and is performed by computational expensive numerical tools such as finite element analysis (FEA). In this respect some studies propose very fast FEA models used in layup optimization. However, the lower bound of the execution time of FEA models is determined by the global linear system solving which in some complex applications can be unacceptable. Moreover, in some situation it may be highly preferred to decrease the optimization time with the cost of a small reduction in the analysis accuracy. In this paper we explore some machine learning techniques in order to estimate the failure of a layup configuration. The estimated response can be qualitative (the configuration fails or not) or quantitative (the value of the failure factor). The procedure consists of generating a population of random observations (configurations) spread across solution space and evaluating using a FEA model. The machine learning method is then trained using this population and the trained model is then used to estimate failure in the optimization process. The results obtained are very promising as illustrated with an example where the misclassification rate of the qualitative response is smaller than 2%.