Earticle

In recent years, Distributed Denial of Service (DDoS) attacks have become increasingly frequent, posing serious threats to the security and stability of network systems. To enhance the effectiveness of DDoS detection, this paper proposes a deep learning model that integrates Convolutional Neural Networks (CNN) with a Transformer architecture to achieve efficient recognition of multiple types of attacks on the CIC-DDoS2019 dataset. By combining feature extraction and temporal modeling, the model fully captures both spatial and contextual information in network traffic, significantly improving detection accuracy and robustness. Experimental results demonstrate that the proposed method outperforms traditional CNN-based models across several sub-datasets, achieving higher accuracy, recall, and F1 scores while maintaining a favorable balance between training and inference time. This research offers new insights and technical support for developing efficient and scalable intelligent network defense systems.

In recent conflicts, such as those in Ukraine and the Middle East, the critical importance of advanced combined air-defense missile systems for combat effectiveness has become increasingly apparent. These integrated systems are tasked not only to detect and engage a spectrum of enemy threats but also to adapt swiftly and reliably to dynamic, rapidly changing circumstances, especially after missile launch. To verify advanced combined air-defense missile system's functionality after missile launch, it is ideal to conduct comprehensive system tests using various scenarios and actual flight tests. However, due to limitations such as high cost and lengthy setup time, this approach is often difficult to implement in reality. In the past, system verification after missile launch has mainly relied on component-level inspections and HILS tests. While these methods can partially check individual system functions, they are not sufficient to fully verify all system functions as a whole. This paper proposes a system integration test method that uses a system integration simulator capable of simulating various combat scenarios and a missile beacon equipped with a missile flight model. Missile flight model uses the same algorithm as the guidance control method employed in actual missiles. Through the adoption of the proposed system integration test method, it was confirmed that a majority of previously limited and difficult-to-verify missile engagement-related test items can now be comprehensively validated. Furthermore, this method allows for comprehensive verification of the system's performance after missile launch, including its interoperability and response capability from a systemic perspective. It also contributes to increasing the reliability of the combined air-defense missile systems.

The purpose of this study is to explore public perceptions, core themes, and semantic structures surrounding the Korea Billiards Federation (KBF) using big data analysis. As the KBF transitions from a traditional recreational organization to a structured sports industry leader, it is essential to understand how the public discusses and associates the organization across digital platforms. To achieve this, data were collected from Naver, Daum, and Google between April 18, 2022, and November 30, 2024, using the keyword "Korea Billiards Federation." A total of 15,758 cases were gathered and analyzed through the Textom platform. After preprocessing the text, keyword extraction was performed based on Term Frequency (TF) and Term Frequency-Inverse Document Frequency(TF-IDF). Subsequently, a semantic network analysis was conducted using UCINET 6.0 and NetDraw to examine degree, closeness, and betweenness centrality. A CONCOR(Convergence of Iterated Correlations) analysis was also applied to classify clusters and interpret contextual meaning. As a result, four key semantic clusters emerged. First, the Division League cluster focuses on the KBF’s role in grassroots and amateur league operations. Second, the Federation cluster reflects governance, national tournaments, and referee systems. Third, the Professional cluster highlights the structure of professional leagues, rankings, and prize systems. Fourth, the Achievement cluster represents individual performance, championships, and player branding. These findings provide strategic insights into the federation’s branding, policy development, and marketing directions and serve as a foundation for the systematic development of Korea’s billiards industry.

Road infrastructure is a key facility that connects economy, logistics, and life, and maintenance is essential for traffic safety. Damage such as potholes and cracks increases accidents and costs, related complaints and damages are rapidly increasing in Korea. To solve this problem, this study proposed a model that automatically detects road damage such as potholes and cracks using the YOLOv8n model for real-time detection. The experiment was conducted based on Roboflow's RoadDamages Detection dataset and the collected dataset. The proposed model was designed with the aim of high precision to Detection of Road Damage(DRD) in real-time, and Accuracy 0.83 and FPS 25 per second were achieved through data augmentation and optimized hyperparameters. By utilizing this, it can increase road maintenance efficiency, contribute to automation of road management and cost reduction, reduce traffic accidents, and strengthen in terms of traffic safety and economy. In addition, it can be expected to be used in various fields such as traffic infrastructure management. In the future, it plans to improve detection accuracy and speed through various backbone integration, data expansion, and hardware optimization.

A Recently AI systems have increasingly focused on integration with various systems for classification and recognition, including IoT applications. This paper introduced to integrate to speech recognition and object detection for user recognition system. The speech recognition model incorporates preprocessing techniques based on voice signal processing, utilizing features such as Mel spectrogram, Mel-frequency Cepstral Coefficients (MFCC), and chroma. These signal processing was important in recently speech recognition research field. also, it can be able to makes elaborate to word classification. so ours model was consist of Convolutional Neural Network(CNN) based model. according to CNN model was simple architecture, it was used to low memory and high inference time. The chroma analysis was consist of voice Pitch data. So, we can classifier to user gender using this analysis. The Your Only Look Once(YOLO) object-based detection research has been actively conducted recently. this model has low memory, high inference speed and great performance accuracy. ours system has integrate to word classification, gender classification and YOLO object detection system. this system worked in user authentication in the administrator system. the user vocalize a word to issue a simple command, and the user’s voice pattern and characteristics are classified, and the gender classification system classifies the gender after determining the voice pitch for further user recognition. Finally we used the QT framework to construct applications and fuse systems to make them easily accessible to users.

We present a systematic review of paradigm shifts in computer vision from 2020 to 2025. The survey centers on Vision Transformers(ViT), large-scale self-supervised learning contrastive, MAE/BEiT, multimodal pretraining CLIP, SAM, diffusion-based generation, and 3D representations via NeRF. Using a literature-synthesis framework, we compare architectures, training regimes, and transfer benefits and limits across major tasks. Evidence shows transformer families rival or surpass CNNs on dense-prediction task detection, segmentation, while diffusion models enable stabler training and higher-quality generation than GANs. Self-supervised learning reduces labeling cost and improves generalization in low-label regimes. Multimodal models unlock zero-shot and open-vocabulary recognition; foundation models such as SAM demonstrate general-purpose segmentation. Persisting challenges include data bias, substantial compute/energy demand, and limited explainability. We recommend efficiency-oriented compression distillation, pruning, quantization, green-AI practices, and guidelines for responsible use of foundation models. The outlook highlights edge/embedded realtime vision, 3D/video understanding, and applications in healthcare, remote sensing, and AR/metaverse. Overall, the period is defined by large-scale pretraining, a shift to transformers, multimodal integration, and advances in 3D—pointing to the next goal: responsible and efficient vision AI.

In recent years, desktop and gantry-type robots have been widely adopted across industrial applications, ranging from logistics and 3D printing to factory automation and smart farming. Conventional systems, often driven by AC servo motors and configured with ball screw-based linear motion guides, face limitations such as mechanical friction, slip, and energy loss due to contact based motion stages. To address these challenges, this study proposes an advanced position control framework for gantry-type collaborative robots by integrating a digital PID (Proportional-Integral-Derivative) controller with AI based parameter auto tuning techniques. The proposed approach leverages machine learning and reinforcement learning algorithms to optimize control parameters dynamically, overcoming the limitations of traditional manual tuning. An Octave based simulation environment is used to validate initial PID settings, which are subsequently applied to a real world gantry-type robot platform. The results demonstrate significant improvements in real time control accuracy, trajectory tracking, and operational stability. This research not only enhances precision and responsiveness in collaborative robot systems but also introduces a generalized control methodology applicable to various automation devices utilizing AC servo motors. The AI-augmented control framework shows strong potential for deployment in diverse fields such as manufacturing, assembly, inspection, logistics, and medical robotics. Ultimately, the study contributes to establishing a scalable and intelligent control standard for next-generation industrial robotic systems.

The growing demand for electric vehicles (EVs) has led to a surge in retired lithium-ion batteries, highlighting the need for efficient second-life battery management. This study presents an integrated battery reuse validation framework developed by the Korea Automotive Technology Institute (KATECH), combining real-time vehicle data and offline diagnostic testing. The framework utilizes two complementary subsystems: the State Estimation Platform (SEP) and the State Measurement Platform (SMP), which jointly assess battery health through online monitoring and electrochemical testing. These results are further processed by the Grade Classification Platform (GCP) and State Prediction Platform (SPP) to determine reuse suitability and predict remaining useful life (RUL). In a case study of 500 used battery packs, the framework achieved a classification accuracy of 93.2% against expert assessments, and its RUL predictions showed an average error margin of ±7.5% over a six-month field deployment. The platform's automated, data-driven approach enhances safety, performance, and economic viability of second-life batteries. This work supports scalable, intelligent battery reuse ecosystems aligned with circular economy principles.

Diabetes mellitus represents a significant global health challenge requiring accurate early prediction and transparent clinical decision-making tools. While traditional machine learning models achieve high predictive accuracy, their "black-box" nature limits clinical adoption due to lack of interpretability. We developed an ensemble model combining Random Forest, XGBoost, and Logistic Regression using soft voting classification on the Pima Indians Diabetes Dataset. Data preprocessing included Synthetic Minority Oversampling Technique (SMOTE) to address class imbalance and feature standardization. Model explanations were generated using LIME and SHAP, which were subsequently processed by GPT-3.5-turbo to produce natural language clinical interpretations for individual patient predictions. Our hybrid approach successfully bridges the gap between machine learning accuracy and clinical interpretability. The framework demonstrates significant potential for real-world clinical deployment by providing both accurate predictions and comprehensible explanations, thereby supporting evidence-based diabetes care and improving patient outcomes. The core contribution of this study is not merely improving prediction accuracy, but proposing a novel explainable framework that integrates XAI techniques with large language models to generate natural language clinical interpretations that are easily understood by both healthcare professionals and patients.

Accurate and continuous environmental monitoring is a key element in implementing smart cities to protect civic health, optimize urban services, and make data-driven policy decisions. However, real-time, city-wide measurement of multidimensional environmental indicators such as CO₂, PM2.5, and VOCs requires the installation of large-scale physical sensors, which poses practical limitations such as cost, maintenance, and spatial constraints. To address this issue, this study proposes a multi-target regression-based soft sensor framework that simultaneously predicts multiple environmental indicators using readily available auxiliary data in cities, such as traffic volume, weather information, and population density. Using techniques such as Random Forest, LightGBM, and Multi-Output Regressor, we construct an integrated prediction model that considers the correlation between various output variables. Even in areas with limited monitoring stations, we achieve an average R² of over 0.80. The proposed model can be integrated with smart city public services, environmental policies, and real-time alert systems to enhance the efficiency and responsiveness of urban environmental management.

This paper proposes a real-time object detection system for railway safety using the Fast YOLO deep learning framework. Using a dataset of over 10,000 annotated images captured from onboard cameras, the system detects people, animals, and obstacles on railway tracks under various environmental conditions. Preprocessing methods including background subtraction and Gaussian modeling enhance detection robustness, achieving 15% relative improvement over baseline in low-light conditions. Experimental results demonstrate high precision (0.925 for people, 0.893 for animals, 0.878 for obstacles) with real-time processing at 38 FPS on NVIDIA GTX 1080 GPU. Our Fast YOLO implementation outperforms Faster R-CNN by 3.2- fold in speed while maintaining comparable accuracy (mAP of 0.84 vs 0.86) and surpasses SSD by 8.7% in detection accuracy. The system achieves 95.2% detection rate for stationary hazards and 91.6% for moving objects, with false positive rates below 2.3%. Field tests over 6 months demonstrated 99.7% uptime reliability and successful prevention of 12 potential incidents. The findings confirm Fast YOLO's effectiveness for automated railway safety monitoring, providing a practical solution for real-world deployment.

This study empirically analyzed the effects of internship job satisfaction on career decision-making and turnover intention among culinary arts students at university, while also incorporating big data–based text analysis to explore the potential for convergence between social science and data science research. A survey was conducted with 500 culinary arts students across universitys nationwide, with respondents consisting of 87% male and 13% female students. The collected data were analyzed using IBM SPSS Statistics and AMOS through descriptive statistics, correlation analysis, regression analysis, and structural equation modeling (SEM), with mediation effects also tested. In addition, text mining and sentiment analysis were performed on internship review texts collected from online job portals and blogs to complement the quantitative findings with qualitative insights. The results revealed that internship job satisfaction had a positive effect on career decision-making and a negative effect on turnover intention, while also playing a significant mediating role in the relationship between career decision-making and turnover intention. The text analysis further confirmed that positive keywords such as “learning,” “growth,” and “professionalism” were closely associated with job satisfaction and career decision-making, whereas negative keywords such as “long working hours,” “low pay,” and “hierarchical culture” were strongly related to turnover intention. This study provides practical baseline data for career development and internship program improvement for culinary arts students at university, while also offering academic and practical contributions by presenting a novel methodological framework that integrates social science research with big data analysis.

Effective emotional support conversations demand nuanced, multi-turn interactions that adaptively employ context-sensitive strategies—an area where large language models (LLMs) often fall short despite their strong general capabilities. To address this gap, we propose a multi-task learning framework that jointly fine-tunes a lightweight DialoGPT model to generate supportive responses and predict the support strategy stage. Using uncertainty-based loss weighting, our method dynamically adjusts multi-task learning objectives based on task-specific uncertainty, enabling balanced optimization between generation and classification tasks. Experiments on the psychologically grounded ESConv dataset show significant improvements, achieving an accuracy of 86.4% and a weighted F1 score of 0.86 in the next-stage strategy prediction task, with particularly strong performance in early dialogue phases such as Exploration. Our study demonstrates that compact LLMs, when guided by task-specific supervision, can effectively deliver strategy-aware emotional support, advancing scalable and reliable mental health conversational agents.

Our study analyzed and evaluated the attack performance of hybrid combinations of Gaussian, Salt & Pepper, and Sine wave-based noise attack models. The experimental results confirmed that the hybrid attack model had the greatest impact on classification accuracy. Although some attacks did not change the output scores, misclassification was observed in emotional label classification. Our study validated that Gaussian, Salt & Pepper, and Sine wave-based noise attacks exploit security vulnerabilities that affect the learning of speech recognition systems and demonstrated the threat posed by hybrid attack models. The Gaussian model showed a significant decrease in emotion scores, while other models maintained high emotion scores. In terms of classification accuracy, changes were observed in the Gaussian, salt and pepper, and hybrid Gaussian models. Emotion attacks were demonstrated in the hybrid Gaussian, Salt & Pepper, and Hybrid sine wave experiments.

This paper propose a system to investigate the psychological impact to 3D sound on users. Growing demand for technological approaches to stress relief in the digital environments of modern society, spatial audio technology increasingly apply to various media content as an effective tool for enhancing a sense of presence and immersive experiences. However, there is still a lack of empirical data regarding the physiological effects of spatial audio on users in gaming environments. This study conducted an experiment using identical game content with two different sound settings: spatial audio (5.1 channel) and stereo audio. Based on EEG (electroencephalogram) data collected during game play, the study compared the effects of each sound environment on stress reduction. The experimental results empirically verify the positive impact of spatial audio on psychological stability and provide insights for designing immersive sound environments in games. These findings may help develop more effective audio-based stress reduction strategies and expand spatial sound applications in entertainment and therapy.

The purpose of this study is to improve the usability of toilet paper dispensers in public restrooms. To this end, it analyses user needs by applying the five-stage design thinking model, and thereby proposes a design solution. The researcher surveyed the types and installation locations of toilet paper dispensers in public restrooms at I University in Gyeongsangnam-do and identified user inconveniences through interviews and brainstorming with 25 diverse users. In particular, this study considered the physical limitations of various users, including older adults and individuals with disabilities, and applied the five principles of universal design for citizen convenience spaces in Seoul Metropolitan Government (Cognitiveness, Accessibility, Diversity, Safety, and Sustainability). In the stage ‘Define’, this study analysed users' hidden needs, focusing on the difficulty of controlling the amount of toilet paper, the inconvenience of checking the remaining amount, and hygiene-related concerns. In the stage ‘Ideate’, the study established design directions that considered feasibility and creativity, such as combining an automatic cutting device with a remaining-amount indicator. Through the prototype design using 3D modeling and rendering, it visually realized the applicability of universal design in terms of user convenience, hygiene, resource waste minimization, and maintenance efficiency. This study proposed a design direction that not only practically improves the usability of public restroom toilet paper dispensers but also enhances accessibility for diverse users, including the socially disadvantaged groups.

This paper focuses on the Multi-Reflection Puzzle, which is one of the core puzzles in the 3D multiplayer game, Isle of the Four Seasons. The paper analyzes the puzzle's effects on users' cognitive functions based on eye tracking. Users perform a cognitive activity requiring repeated distraction and refocusing to solve the puzzle, which involves predicting the reflection path of a laser. Real-time gaze data were recorded with a Python webcam eye tracker during gameplay, and ROI analysis with gaze trajectory visualization provided indicators of gaze concentration, attention switching, and blink frequency. The results revealed high gaze focus on the puzzle center region and gaze trajectory patterns resembling reflection paths in certain users. This suggests that reflection puzzles stimulate strategic visual reasoning and problem-solving processes. Additionally, differences in switching frequency and blink rate among participants reflected variations in strategic approaches and engagement levels. This study demonstrates the potential for real-time cognitive flow analysis and immersion assessment using eye tracking, a limitation of existing puzzle studies. This method can inform the design of game-based cognitive stimulation and feedback systems.

This study investigated the influence of scene composition on narrative pacing through a quantitative analysis of two animated short films, Farewell and One Small Step. The analysis was grounded in theories of narrative time, including Bergson’s notion of durée, Husserl’s triple temporal structure, and Deleuze’s distinction between movement-image and time-image. Editing variables such as shot length, cut frequency, and rhythm were found to shape viewers’ emotional engagement, and Cinemetrics analysis identified distinct pacing patterns with measurable psychological effects. The findings were further situated within broader temporal theories from physics (Einstein, Minkowski) and phenomenology (Augustine, Kant). We present a new methodological framework that bridges perceptual time and narrative structure in audiovisual storytelling, advancing the quantitative study of audiovisual time. This framework establishes a model for analyzing audiovisual time not only in traditional films but also in emerging media environments, providing a foundation for future research at the intersection of narrative theory, cinematic practice, and media technology.

With the rise of short-form content and the growing integration of generative AI (Artificial Intelligence Generated Content, AIGC) technologies in drama production, this study extends the AI Device Use Acceptance (AIDUA) model to examine the factors influencing acceptance and rejection of AIGC dramas based on Chinese traditional mythology. Drawing on survey data from 528 Chinese viewers and employing structural equation modeling, the study identifies six key factors—social influence, hedonic motivation, perceived anthropomorphism, artistic quality, novelty, and cultural traditionality—as significant predictors of performance expectancy. Emotional responses, shaped by performance expectancy, were found to differentially influence acceptance and rejection intentions. The findings underscore the importance of emotional appraisal in user responses to AIGC content and provide empirical insights for the digital transformation of traditional intellectual property (IP) and the strategic planning of culturally resonant AIgenerated content.

This study proposes a solution to the challenges encountered in Maya's HumanIK system when creating animations that require a character’s Effector to interact with their own body parts, such as the waist or knees. In conventional workflows, using parentConstraint often fails when a character’s hand or foot must be connected to their own body, due to conflicts with the character rig structure. To address this issue, the proposed method utilizes a Rivet node to track the motion of a specific body part. The positional data from the Rivet is transferred to a separate locator, which is then used to recalculate and reassign the position of the Effector, thereby enabling stable control over the hand’s attachment or constraint. To enhance workflow efficiency, a custom automation script was developed, and its stability was verified through testing. The proposed method shows potential for enabling more flexible and robust representations of interactions between self-body contact.

Jump displacement is regarded as part of instruction displacement, which can be used for code diversification and code randomization. In this paper, we implement jump displacement technique on benign PE (Portable Executable) files through IDA Pro 7.0 in Windows systems. We then evaluate the performance of Jump displacement technique. More specifically, we implement jump displacement technique on 30 benign PE files and perform evaluation of them through IDA Pro 7.0. Our evaluation results demonstrate that jump displacement can cause an increase in the number of anti-virus engines misclassifying benign PE files as malware. Moreover, we discern that the larger number of machine codes affected by jump displacement technique in benign PE files can lead to the higher chance that IDA Pro 7.0 fails to disassemble correctly benign PE files, leading to malfunction of benign PE files.

Our research focuses on the possibility of adversarial attacks that manipulate subtle audio signals to influence model predictions, considering the current situation where audio alterations expose security vulnerabilities. Against this backdrop, it is necessary to evaluate the impact of existing audio attacks on their detectability even under realistic physical attack conditions. In particular, our research centers on low-perceptibility attacks—such as Fade in/out, Sine mod, Bit reduction, and Hybrid attacks—that induce errors in speech recognition learning without affecting the human auditory system, aiming to identify their attack effectiveness. In the research methodology, we conducted experiments on six types of attack methods and analyzed the detection evasion capabilities and attack efficiency of the models through quantitative results, including attack success rates, prediction confidence, and visual similarity. The results revealed that recognition rates decreased in Fade in/out-based attacks, and Sine mod attacks induced emotion recognition error rates without degrading audio quality. Additionally, the Fade+ Sine modulation + Bit reduction based Hybrid attack model demonstrated the most balanced distortion and was confirmed to be the most successful in evading detection.

This paper investigates the narrative ending structures of six major South Korean films about the Korean War produced since 2000, a period in which the genre has shifted from Cold War anti-communist ideology to a perspective framing the conflict as a national tragedy. As the Korean War remains a historically unresolved event, concluded by an armistice rather than a peace treaty, this study focuses on how these films strategically employ narrative closure. Using the theoretical frameworks of Roland Barthes’s narrative codes (hermeneutic and proairetic) and Richard Neupert’s typology of film endings, this paper analyzes the micro-level mechanics and macro-level structures of cinematic closure. The analysis reveals two opposing strategies. The first is the closed textstructure, seen in films like Taegukgi: The Brotherhood of War, 71: Into the Fire, and Operation Chromite. These films provide ideological certainty and emotional catharsis by framing tragic sacrifices within a nationalistic context or constructing heroic myths, thereby offering a definitive meaning to the historical trauma. The second strategy is the open story structure, employed by films such as Welcome to Dongmakgol, The Front Line, and The Battle of Jangsari. Through ambiguity, unresolved questions, and nihilistic circularity, these narratives critique the absurdity of war, challenge official histories, and leave the ethical task of memory and commemoration to the audience. Ultimately, this paper contends that these ending structures are not mere aesthetic choices but active strategic devices that reflect South Korean society’s complex and ongoing negotiation with its unresolved national trauma, historical memory, and ideological conflicts.

In today’s rapidly changing educational environment, there is an increasing demand for a flexible and repeatable system that supports the full cycle of curriculum development, operation, assessment, and continuous improvement. The widely used Plan-Do-Check-Act (PDCA) model has shown structural limitations, particularly in providing real-time feedback, agile improvement, and data-driven operations. To address these challenges, this study proposes EduOps, an educational operations model inspired by the core philosophy of DevOps. EduOps consists of six stages - Plan, Design, Deploy, Operate, Monitor, and Improve - and features a modular architecture that integrates curriculum planning, automated course delivery (CI/CD), performance monitoring, and AI-based feedback tracking. Unlike prior uses of the term “EduOps,” which largely referred to IT service or educational technology operations, this study defines EduOps as a higher-education-centered lifecycle management system that technologically unifies curriculum design, operation, and feedback in a continuous loop. A prototype implementation was developed and applied to actual courses, with effectiveness validated through both quantitative indicators (student performance, engagement metrics) and qualitative evaluations (faculty interviews, learner feedback). Results demonstrate that EduOps enhances agility and transparency in classroom operations, supports continuous outcome-based improvement, and provides a foundation for data-driven innovation. This research establishes EduOps as a distinct, replicable model for Education 4.0 and demonstrates its practical viability in higher education.

This study addresses the need to understand brand success by analyzing a crucial source of customer feedback: online reviews. We designed a comparative analysis to investigate the different factors that drive the success of two major fast-food brands, BBQ and KFC, in the competitive market. To achieve this, we combined several advanced analytical techniques, including text mining, sentiment analysis, TF-IDF, and CONCOR network analysis, to systematically process and evaluate their respective customer review data. Our results from the sentiment analysis showed that both brands garnered strong positive feedback, progressing from 'satisfaction' to 'recommendation' and 'best.' The TF-IDF analysis successfully identified key keywords for each brand, while the CONCOR network analysis revealed distinct competitive clusters: 'quality, localization, and service' for BBQ and 'price, convenience, and global' for KFC. These findings provide valuable insights into brand strategy. The effects of this study are the proposal of four major success factors— advanced localization, quality control, data capability strengthening, and local partnerships—which offer a strategic framework for businesses operating in this industry.

This study compares visual communication design outputs created through two generative AI platforms— ChatGPT and Midjourney—within an educational context. Focusing on three design categories (poster, logo, and package), the research examines how each platform supports the creative process. ChatGPT, as a language-based model, showed strengths in generating structured ideas and visual direction, particularly for logo where branding and concept clarity are essential. In contrast, Midjourney, as an image-generating AI, excelled in producing a wide range of artistic and expressive visuals, making it especially effective for poster and package design that benefits from painterly and atmospheric elements. Evaluation criteria were based on the specific standards pursued in poster, logo, and package design to effectively assess the generated outputs. The study concludes that each platform offers unique advantages depending on the design task, suggesting that their complementary use can enhance both design education and professional practice. These findings highlight the importance of platform literacy in future visual communication curricula and propose a hybrid approach to AI-assisted design.

This study aims to examine how Artificial Intelligence (AI) can be practically applied to post-production in broadcasting and to evaluate its implications for future content creation. To this end, it focuses on the case of Earth Sweepers, a 2024 MBC reality variety program recognized as the first terrestrial TV show in Korea to adopt AI-based editing and de-identification under outdoor filming conditions. The research method involved analyzing the integration of two key technologies: automatic editing developed by the Electronics and Telecommunications Research Institute (ETRI) and de-identification technology developed by the Korea Electronics Technology Institute (KETI). Both tools were implemented as plug-ins for Adobe Premiere Pro to ensure seamless use within established editing workflows. The results revealed that AI significantly reduced production time—overall editing was shortened by approximately 35–40%, and de-identification achieved over 90% accuracy while cutting costs by more than 20 million KRW per episode. AI-assisted editing also improved multicamera alignment, scene segmentation, and highlight extraction, while de-identification enhanced both visual quality and compliance with regulatory requirements. The findings highlight the potential of human-AI collaboration in broadcasting and underscore the importance of establishing standardization, professional training, and institutional support to ensure sustainable adoption. This research provides valuable insights for guiding innovation and shaping the future ecosystem of AI broadcasting.

This paper explores the functional, historical, and artistic significance of music and sound in video games including computer g. As the gaming industry has evolved from its early technological limitations into a sophisticated form of interactive media, audio has emerged as a central element in shaping player immersion and emotional engagement. Game sound, encompassing both background music and sound effects, is no longer a supplementary feature but a fundamental component of modern game design. The study examines how audio elements function interactively in games, particularly in enhancing realism, providing feedback, and supporting narrative progression. The historical development of game audio is traced from the simple 4-channel mono sounds of early consoles to the high-fidelity, cinematic soundscapes of contemporary titles. Special emphasis is placed on the importance of repetition in game music composition, distinguishing it from commercial music. The paper also analyzes genre-specific approaches to game music, highlighting the use of orchestral, electronic, rock, hip-hop, and fusion styles tailored to various gameplay experiences. Furthermore, it profiles influential composers—including Koji Kondo, Glenn Stafford, Mick Gordon, Jesper Kyd, and Hans Zimmer—who have significantly shaped the evolution of video game music. The contributions of Korean composer Yang Seung-hyuk are also discussed as part of the growing international presence of Korean game soundtracks. In conclusion, this study underscores that music and sound are indispensable to the game development process, enhancing not only the emotional and narrative depth of games but also their artistic value. As video games continue to integrate with other forms of media, the role of audio will remain vital in defining the future of interactive entertainment.

This study analyzes AI education policies in South Korea and Japan between 2017 and 2024, a period when the necessity of AI education gained increasing prominence. Both countries recognize AI education as strategically important, yet their policies diverge in underlying assumptions, orientations, and implementation trajectories. South Korea advances a strategy-driven, technology-oriented approach, positioning AI education as an instrument for economic growth and global competitiveness through sequential reforms such as national roadmaps, curriculum revisions, and digital textbooks. In contrast, Japan pursues a vision-driven, holistic trajectory under the framework of Society 5.0, embedding AI education within a broader ethical and socially oriented context. These contrasting assumptions shape how each country defines core competencies, structures curricula, and conceptualizes the role of AI in society. South Korea emphasizes technical skills such as coding, AI literacy, and data science, whereas Japan foregrounds social and ethical dimensions, including digital citizenship and the early integration of AI ethics. Taken together, the findings underscore the need for AI education policies that balance technical, cognitive, social, and ethical competencies. Ultimately, this comparative analysis contributes to the policy discourse of AI education by demonstrating how distinct national approaches can offer complementary insights for future-oriented AI education strategies.

The evolution of generative artificial intelligence—exemplified by OpenAI's ChatGPT and other large language and vision models—has sparked considerable change across multiple domains, including education. Particularly within immersive media education and digital content production, generative AI introduces new pedagogical possibilities that transcend traditional workflows. This study investigates how generative AI technologies can be practically integrated into projection mapping education, aiming to enrich learner engagement and foster immersive creative experiences in both technical and artistic dimensions. This research centers on formulating an instructional framework that embeds generative AI tools into the stages of projection mapping production: from ideation and conceptual development to asset creation, spatial mapping, and real-time visual synchronization. Students engaged in this process benefit from AI-assisted content generation, including text-to-image modeling, automated storyboard creation, and simulation-based visual testing, which together accelerate their creative pipeline and lower the cognitive load associated with complex technical execution. Moreover, the study explores how the integration of AI technologies influences learners’ sense of immersion, autonomy, and creative confidence. Drawing on a combination of structured instructional design and exploratory field experiments, the project collected both qualitative and quantitative data through observational analysis, participant interviews, and expert validation. Results indicate that AI-supported education not only improves students’ task efficiency and comprehension but also amplifies their emotional and cognitive engagement during production-based activities. In addressing the challenges of educational accessibility and the gap between conceptual learning and realworld production, the study offers a scalable model adaptable to diverse educational environments, including major learners and interdisciplinary design programs. Furthermore, it provides concrete pedagogical strategies for educators seeking to integrate AI technologies into their teaching, with a focus on maximizing learner immersion, creative expression, and project-based outcomes. This case study ultimately contributes to the discourse on next-generation educational practices in immersive content creation, proposing generative AI as a transformative catalyst for both teaching innovation and learner-centered design in the age of intelligent media.