• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.10
• /
• pp.437-446
• /
• 2022

Text classification task from Natural Language Processing (NLP) combined with state-of-the-art (SOTA) Machine Learning (ML) and Deep Learning (DL) algorithms as the core engine is widely used to detect and classify voice phishing call transcripts. While numerous studies on the classification of voice phishing call transcripts are being conducted and demonstrated good performances, with the increase of non-face-to-face financial transactions, there is still the need for improvement using the latest NLP technologies. This paper conducts a benchmarking of Korean voice phishing detection performances of the pre-trained Korean language model KoBERT, against multiple other SOTA algorithms based on the classification of related transcripts from the labeled Korean voice phishing dataset called KorCCVi. The results of the experiments reveal that the classification accuracy on a test set of the KoBERT model outperforms the performances of all other models with an accuracy score of 99.60%.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.1
• /
• pp.63-70
• /
• 2023

Due to the spread of high-definition black boxes and the introduction of mobile applications such as 'Smart Citizens Report' and 'Safety Report', the number of public interest reports for violations of Traffic Law has increased rapidly, resulting in shortage of police personnel to handle them. In this paper, we describe the development of a system that can automatically detect lane violations which account for the largest proportion of public interest reporting videos for violations of traffic laws, using deep learning algorithms. In this study, a method for recognizing a vehicle and a solid line object using a YOLO model and a Lanenet model, a method for tracking an object individually using a deep sort algorithm, and a method for detecting lane change violations by recognizing the overlapping range of a vehicle object's bounding box and a solid line object are described. Using this system, it is expected that the shortage of police personnel in charge will be resolved.

• KIPS Transactions on Software and Data Engineering
• /
• v.11 no.6
• /
• pp.245-254
• /
• 2022

Proteins are the basic unit of all life activities, and understanding them is essential for studying life phenomena. Since the emergence of the machine learning methodology using artificial neural networks, many researchers have tried to predict the function of proteins using only protein sequences. Many combinations of deep learning models have been reported to academia, but the methods are different and there is no formal methodology, and they are tailored to different data, so there has never been a direct comparative analysis of which algorithms are more suitable for handling protein data. In this paper, the single model performance of each algorithm was compared and evaluated based on accuracy and speed by applying the same data to CNN, LSTM, and GRU models, which are the most frequently used representative algorithms in the convergence research field of predicting protein functions, and the final evaluation scale is presented as Micro Precision, Recall, and F1-score. The combined models CNN-LSTM and CNN-GRU models also were evaluated in the same way. Through this study, it was confirmed that the performance of LSTM as a single model is good in simple classification problems, overlapping CNN was suitable as a single model in complex classification problems, and the CNN-LSTM was relatively better as a combination model.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.1
• /
• pp.471-476
• /
• 2024

With the spread of various smart devices and computing devices, big data generation is occurring widely. Machine learning is an algorithm that performs reasoning by learning data patterns. Among the various machine learning algorithms, the algorithm that attracts attention is deep learning based on neural networks. Deep learning is achieving rapid performance improvement with the release of various applications. Recently, among deep learning algorithms, attempts to analyze data using graph structures are increasing. In this study, we present a graph generation method for transferring to a deep learning network. This paper proposes a method of generalizing node properties and edge weights in the graph generation process and converting them into a structure for deep learning input by presenting a matricization We present a method of applying a linear transformation matrix that can preserve attribute and weight information in the graph generation process. Finally, we present a deep learning input structure of a general graph and present an approach for performance analysis.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.8
• /
• pp.59-65
• /
• 2024

In this paper, we present a GPU-based acceleration of computationally intensive self-collision processing in triangular mesh-based cloth simulation. For Compute Unified Device Architecture (CUDA)-based parallel optimization, we propose 1) an efficient way to build, update, and traverse the Bounding Volume Hierarchy (BVH) tree on the GPU, and 2) optimize the Representative-Triangle (R-Triangle) technique on the GPU to minimize primitive collision checking in triangular mesh-based cloth simulations. As a result, the proposed method can handle self-collisions and object collisions of cloth simulation in GPU environment faster and more efficiently than CPU-based algorithms, and experiments on various scenes show that it can achieve simulation results that are 5x to 10x faster. Since the proposed method is optimized for BVH on GPU, it can be easily integrated into various algorithms and fields that utilize BVH.

• Journal of Internet Computing and Services
• /
• v.21 no.4
• /
• pp.109-116
• /
• 2020

With the advent of high-speed railways, railways are one of the most frequently used means of transportation at home and abroad. In addition, in terms of environment, carbon dioxide emissions are lower and energy efficiency is higher than other transportation. As the interest in railways increases, the issue related to railway safety is one of the important concerns. Among them, visual abnormalities occur when various obstacles such as animals and people suddenly appear in front of the railroad. To prevent these accidents, detecting rail tracks is one of the areas that must basically be detected. Images can be collected through cameras installed on railways, and the method of detecting railway rails has a traditional method and a method using deep learning algorithm. The traditional method is difficult to detect accurately due to the various noise around the rail, and using the deep learning algorithm, it can detect accurately, and it combines the two algorithms to detect the exact rail. The proposed algorithm determines the accuracy of railway rail detection based on the data collected.

• Journal of KIISE
• /
• v.44 no.5
• /
• pp.476-485
• /
• 2017

The tactical operation centers of future anti-aircraft missile systems provide the environment for the research on future air threats, tactical information, integrated battlefield environment creation and management, engagement control and command and control algorithms. To develop the key functional elements of integrated battlefield situation creation and processing and tactical operation automation processing operations, battery/battalion tactical operation control and reconfiguration design software are required. Therefore, the algorithm software of each function and the tactical operation display software and link software for interworking between equipment were developed as reconfigurable through a data-centric design. In this paper, a tactical operation display framework that can be reconfigured on the operation display of the tactical operations according to the battery/battalion is introduced. This tactical operation display framework was used to develop a common data model design for the reconfigurable structure of multi-role tactical operations with battery / battalion and mission views, and a display configuration tool that provides a tactical operation display framework for view development was also developed using the MVC pattern. If the tactical operation display framework is used, it will be possible to reuse the view design through the common base structure, and a view that can be reconfigured easily and quickly will also be developed.

• International Journal of Contents
• /
• v.11 no.1
• /
• pp.41-51
• /
• 2015

Management systems for electronic library have been developed on the basis of Client/Server or ASP framework in domestic market for a long time. Therefore, both service provider and user suffer from their high cost and effort in management, maintenance, and repairing of software as well as hardware. Recently in addition, mobile devices like smartphone and tablet PC are frequently used as terminal devices to access computers through the Internet or other networks, sophisticatedly customized or personalized interface for n-screen service became more important issue these days. In this paper, we propose a new scheme of integrated management system for electronic library based on SaaS and Web Standard. We design and implement the proposed scheme applying Electronic Cabinet Guidelines for Web Standard and Universal Code System. Hosted application management style and software on demand style service models based on SaaS are basically applied to develop the management system. Moreover, a newly improved concept of duplication check algorithm in a hierarchical evaluation process is presented and a personalized interface based on web standard is applied to implement the system. Algorithms of duplication check for journal, volume/number, and paper are hierarchically presented with their logic flows. Total framework of our development obeys the standard feature of Electronic Cabinet Guidelines offered by Korea government so that we can accomplish standard of application software, quality improvement of total software, and reusability extension. Scope of our development includes core services of library automation system such as acquisition, list-up, loan-and-return, and their related services. We focus on interoperation compatibility between elementary sub-systems throughout complex network and structural features. Reanalyzing and standardizing each part of the system under the concept on the cloud of service, we construct an integrated development environment for generating, test, operation, and maintenance. Finally, performance analyses are performed about resource usability of server, memory amount used, and response time of server etc. As a result of measurements fulfilled over 5 times at different test points and using different data, the average response time is about 62.9 seconds for 100 clients, which takes about 0.629 seconds per client on the average. We can expect this result makes it possible to operate the system in real-time level proof. Resource usability and memory occupation are also good and moderate comparing to the conventional systems. As total verification tests, we present a simple proof to obey Electronic Cabinet Guidelines and a record of TTA authentication test for topics about SaaS maturity, performance, and application program features.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.48 no.3
• /
• pp.62-70
• /
• 2011

Nowadays image sensor is an essential component in many multimedia devices, and it is covered by a color filter array to filter out specific color components at each pixel. We need a certain algorithm to combine those color components reconstructed a full color image from incomplete color samples output from an image sensor, which is called a demosaicking process. Most existing demosaicking algorithms are developed for ideal image sensors, but they do not work well for the practical cases because of dissimilar characteristics of each sensor. In this paper, we propose a new demosaicking algorithm in which the color filter characteristics are fully utilized to generate a good image. To demonstrate significance of our algorithm, we used a commerically available sensor, CBN385B, which is a sort of Honeycomb-style CFA(Color Filter Array) CCD image sensor. As a performance metric of the algorithm, PSNR(Peak Signal to Noise Ratio) and RGB distribution of the output image are used. We first implemented our algorithm in C-language for simulation on various input images. As a result, we could obtain much enhanced images whose PSNR was improved by 4~8 dB compared to the commonly idealized approaches, and we also could remove the inclined red property which was an unique characteristics of the image sensor(CBN385B).Then we implemented it in hardware to overcome its problem of computational complexity which made it operate slow in software. The hardware was verified on Spartan-3E FPGA(Field Programable Gate Array) to give almost the same performance as software, but in much faster execution time. The total logic gate count is 45K, and it handles 25 image frmaes per second.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.37C no.11
• /
• pp.1094-1103
• /
• 2012

Recently, software industry regarding national defense increases system development of distributed simulation system of M&S based to overcome limit of resource and expense. It is one of key technologies for offering of mutual validation among objects and reuse of objects which are discussed for developing these systems. RTI, implementation of HLA interface specification as software providing these technologies uses Federation Object Model for exchanging information with joined federates in the federation and each federate has a characteristic that is supposed to have identical FOM in the federation. This technology is a software which is to provide the core technology which was suggested by the United state's military M&S standard framework. Simulator, virtual simulation, and inter-connection between military weapons system S/W which executes on network which is M&S's core base technology, and it is a technology which also can be used for various inter-connection between S/W such as game and on-line phone. These days although RTI is used in military war game or tactical training unit field, there is none in Korea. Also, it is used in mobile-game, distribution game, net management, robot field, and other civilian field, but the number of examples are so small and informalized. Through this developing project, we developed the core technique and RTI software and provided performance of COTS level to improve communication algorithms.