• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.2
• /
• pp.311-319
• /
• 2017

Recently invasion of privacy problem in medical information have been issued following the interest in secondary use of large medical information. These large medical information is very useful information that can be used in various fields such as disease research and prevention. However, due to the privacy laws such as Privacy Act and Medical Law, these informations including patients or health professionals' personal information are difficult to utilize secondary. Accordingly, various methods such as k-anonymity, l-diversity and differential-privacy that can be utilized while protecting privacy have been developed and utilized in this field. In this paper, we study differential privacy processing procedure, one of various methods, and find out about the differential privacy problem using Laplace noise. Finally, we propose a new method using the Shamir's secret sharing method and symemetric key encryption algorithm such as AES for this problem.

• The Journal of the Acoustical Society of Korea
• /
• v.27 no.5
• /
• pp.256-261
• /
• 2008

In this paper, we propose a novel approach to improve the performance of minima controlled recursive averaging (MCRA) which is based on the conditional maximum a posteriori criterion. A crucial component of a practical speech enhancement system is the estimation of the noise power spectrum. One state-of-the-art approach is the minima controlled recursive averaging (MCRA) technique. The noise estimate in the MCRA technique is obtained by averaging past spectral power values based on a smoothing parameter that is adjusted by the signal presence probability in frequency subbands. We improve the MCRA using the speech presence probability which is the a posteriori probability conditioned on both the current observation the speech presence or absence of the previous frame. With the performance criteria of the ITU-T P.862 perceptual evaluation of speech quality (PESQ) and subjective evaluation of speech quality, we show that the proposed algorithm yields better results compared to the conventional MCRA-based scheme.

• Journal of the Korean Society for Railway
• /
• v.19 no.6
• /
• pp.833-843
• /
• 2016

Recently, in the field of rail freight transportation, the number of trains dedicated for shippers has been increasing. These dedicated trains, which run on the basis of a contract with shippers, had been restricted to the transportation of containers, or so called block trains. Nowadays, such commodities have extended to cement, hard coal, etc. Most full freight cars are transported by dedicated trains. But, for empty car distribution, the efficiency still remains questionable because the distribution plan is manually developed by dispatchers. In this study, we investigated distribution models delineated in the KTOCS system which was developed by KORAIL as well as mathematical models considered in the state-of-the-art. The models are based on optimization models, especially the network flow model. Here we suggest a new optimization model with a framework of the column generation approach. The master problem can be formulated into a transportation problem with additional constraints. The master problem is improved by adding a new edge between the supply node and the demand node; this edge can be found using a simple shorted path in the time-space network. Finally, we applied our algorithm to the Korean freight train network and were able to find the total number of empty car kilometers decreased.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.43 no.6 s.312
• /
• pp.34-42
• /
• 2006

As art analytic method to uncover interesting patterns hidden under a large volume of data, data mining research has been actively done so far in various fields. However, current state-of-the-arts in data mining research have several challenging problems such as being too ad-hoc. The existing techniques are mostly the ones designed for individual problems, so there is no unifying theory applicable for more general data mining problems. In this paper, we address the problem of classification, which is one of significant data mining tasks. Specifically, our objective is to evaluate radial basis function (RBF) model for classification tasks and investigate its usefulness. For evaluation, we analyze the popular Monk's problems which are well-known datasets in data mining research. First, we develop RBF models by using the representational capacity based learning algorithm, and then perform a comparative assessment of the results with other models generated by the existing techniques. Through a variety of experiments, it is empirically shown that the RBF model has not only the superior performance on the Monk's problems but also its modeling process can be controlled in a systematic way, so the RBF model with RC-based algorithm might be a good candidate to handle the current ad-hoc problem.

• Journal of Internet Computing and Services
• /
• v.15 no.6
• /
• pp.125-132
• /
• 2014

Since frequent pattern mining was proposed in order to search for hidden, useful pattern information from large-scale databases, various types of mining approaches and applications have been researched. Especially, frequent graph pattern mining was suggested to effectively deal with recent data that have been complicated continually, and a variety of efficient graph mining algorithms have been studied. Graph patterns obtained from graph databases have their own importance and characteristics different from one another according to the elements composing them and their lengths. However, traditional frequent graph pattern mining approaches have the limitations that do not consider such problems. That is, the existing methods consider only one minimum support threshold regardless of the lengths of graph patterns extracted from their mining operations and do not use any of the patterns' weight factors; therefore, a large number of actually useless graph patterns may be generated. Small graph patterns with a few vertices and edges tend to be interesting when their weighted supports are relatively high, while large ones with many elements can be useful even if their weighted supports are relatively low. For this reason, we propose a weight-based frequent graph pattern mining algorithm considering length-decreasing support constraints. Comprehensive experimental results provided in this paper show that the proposed method guarantees more outstanding performance compared to a state-of-the-art graph mining algorithm in terms of pattern generation, runtime, and memory usage.

• Journal of Broadcast Engineering
• /
• v.26 no.3
• /
• pp.247-257
• /
• 2021

This paper proposes a new technique for calibrating a multi-view RGB-D camera using a 3D (dimensional) skeleton. In order to calibrate a multi-view camera, consistent feature points are required. In addition, it is necessary to acquire accurate feature points in order to obtain a high-accuracy calibration result. We use the human skeleton as a feature point to calibrate a multi-view camera. The human skeleton can be easily obtained using state-of-the-art pose estimation algorithms. We propose an RGB-D-based calibration algorithm that uses the joint coordinates of the 3D skeleton obtained through the posture estimation algorithm as a feature point. Since the human body information captured by the multi-view camera may be incomplete, the skeleton predicted based on the image information acquired through it may be incomplete. After efficiently integrating a large number of incomplete skeletons into one skeleton, multi-view cameras can be calibrated by using the integrated skeleton to obtain a camera transformation matrix. In order to increase the accuracy of the calibration, multiple skeletons are used for optimization through temporal iterations. We demonstrate through experiments that a multi-view camera can be calibrated using a large number of incomplete skeletons.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.6
• /
• pp.709-718
• /
• 2017

The installation of the elevator ARD(Automatic Rescue Device) system has been forced into law in these days in order to safely rescue passengers during power failure. The configuration of the ARD system consists of energy storage device, power converter and control systems. The EDLC(Electric Double Layer Capacitor) are used as energy storage device for rapid charge/discharge purposes. The power conditioning system (PCS) consists of bi-directional converter, 3-phase converter and control system. The dead-beat control system is adopted for most systems however it requires complex mathematical calculations, the high performance microprocessors are mandatory and thus it can be a cause of high manufacturing cost. In this paper the new control method for average current mode control is presented for simple structure. The control algorithm is applied to the single phase system and then expands to three phase system to meet the sysem requirements. The mathematical modeling using average modeling method is presented and analysed by PSIM computer simulation to verifie the validity of the proposed control methods.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.

• Journal of Science of Art and Design
• /
• v.5
• /
• pp.87-129
• /
• 2003

The purpose of this thesis is to propose a new concept of design of the 21st century, on the basis of the study on the general signification of the structures and the signs of industrial product design, by examining the difference between modern and post-modern design, which is expected to lead the users to different design practice and interpretation of it. The starting point of this study is the different styles and patterns of 'Gestalt' in the post-modern design of the late 20th century from modern design - the factor of determination in industrial product design. That is to say, unlike functional and rational styles of modern product design, the late 20th century is based upon the pluralism characterized by complexity, synthetic and decorativeness. So far, most of the previous studies on design seem to have excluded visual aspects and usability, focused only on effective communication of design phenomena. These partial studies on design, blinded by phenomenal aspects, have resulted in failure to discover a principle of fundamental system. However, design varies according to the times; and the transformation of design is reflected in Design Pragnanz to constitute a new text of design. Therefore, it can be argued that Design Pragnanz serves as an essential factor under influence of the significance of text. In this thesis, therefore, I delve into analysis of the 20th century product design, in the light of Gestalt theory and Design Pragnanz, which have been functioning as the principle of the past design. For this study, I attempted to discover the fundamental elements in modern and post-modern designs, and to examine the formal structure of product design, the users' aesthetic preference and its semantics, from the integrative viewpoint. Also, with reference to history and theory of design my emphasis is more on fundamental visual phenomena than on structural analysis or process of visualization in product design, in order to examine the formal properties of modern and post-modern designs. Firstly, In Chapter 1, 'Issues and Background of the Study', I investigated the Gestalt theory and Design Pragnanz, on the premise of formal distinction between modern and post-modern designs. These theories are founded upon the discussion on visual perception of Gestalt in Germany in 1910's, in pursuit of the principle of perception centered around visual perception of human beings. In Chapter 2, I dealt with functionalism of modern design, as an advance preparation for the further study on the product design of the late 20th century. First of all, in Chapter 2-1, I examined the tendency of modern design focused on functionalism, which can be exemplified by the famous statement 'Form follows function'. Excluding all unessential elements in design - for example, decoration, this tendency has attained the position of the international style based on the spirit of Bauhause - universality and regularity - in search of geometric order, standardization and rationalization. In Chapter 2-2, I investigated the anthropological viewpoint that modern design started representing culture in a symbolic way including overall aspects of the society - politics, economics and ethics, and its criticism on functionalist design that aesthetic value is missing in exchange of excessive simplicity in style. Moreover, I examined the pluralist phenomena in post-modern design such as kitsch, eclecticism, reactionism, hi-tech and digital design, breaking away from functionalist purism of modern design. In Chapter 3, I analyzed Gestalt Pragnanz in design in a practical way, against the background of design trends. To begin with, I selected mass product design among those for the 20th century products as a target of analysis, highlighting representative styles in each category of the products. For this analysis, I adopted the theory of J. M Lehnhardt, who gradated in percentage the aesthetic and semantic levels of Pragnantz in design expression, and that of J. K. Grutter, who expressed it in a formula of M = O : C. I also employed eight units of dichotomies, according to the G. D. Birkhoff's aesthetic criteria, for the purpose of scientific classification of the degree of order and complexity in design; and I analyzed phenomenal aspects of design form represented in each unit. For Chapter 4, I executed a questionnaire about semiological phenomena of Design Pragnanz with 28 units of antonymous adjectives, based upon the research in the previous chapter. Then, I analyzed the process of signification of Design Pragnanz, founded on this research. Furthermore, the interpretation of the analysis served as an explanation to preference, through systematic analysis of Gestalt and Design Pragnanz in product design of the late 20th century. In Chapter 5, I determined the position of Design Pragnanz by integrating the analyses of Gestalt and Pragnanz in modern and post-modern designs In this process, 1 revealed the difference of each Design Pragnanz in formal respect, in order to suggest a vision of the future as a result, which will provide systemic and structural stimulation to current design.

• Korean Journal of Remote Sensing
• /
• v.39 no.5_3
• /
• pp.979-995
• /
• 2023

As wildfires are difficult to predict, real-time monitoring is crucial for a timely response. Geostationary satellite images are very useful for active fire detection because they can monitor a vast area with high temporal resolution (e.g., 2 min). Existing satellite-based active fire detection algorithms detect thermal outliers using threshold values based on the statistical analysis of brightness temperature. However, the difficulty in establishing suitable thresholds for such threshold-based methods hinders their ability to detect fires with low intensity and achieve generalized performance. In light of these challenges, machine learning has emerged as a potential-solution. Until now, relatively simple techniques such as random forest, Vanilla convolutional neural network (CNN), and U-net have been applied for active fire detection. Therefore, this study proposed an active fire detection algorithm using state-of-the-art (SOTA) deep learning techniques using data from the Advanced Himawari Imager and evaluated it over East Asia and Australia. The SOTA model was developed by applying EfficientNet and lion optimizer, and the results were compared with the model using the Vanilla CNN structure. EfficientNet outperformed CNN with F1-scores of 0.88 and 0.83 in East Asia and Australia, respectively. The performance was better after using weighted loss, equal sampling, and image augmentation techniques to fix data imbalance issues compared to before the techniques were used, resulting in F1-scores of 0.92 in East Asia and 0.84 in Australia. It is anticipated that timely responses facilitated by the SOTA deep learning-based approach for active fire detection will effectively mitigate the damage caused by wildfires.