- P-ISSN1738-6764
- E-ISSN2093-7504
- KCI
ISSN : 1738-6764
Vol.22 No.2
Abstract
Temporal Action Segmentation (TAS) assigns an action label to each frame in a sequence. While TAS has been extensively studied in contexts such as daily activities, sports, and surgical videos, its application to continuous dance motion derived from video-based skeleton data is still limited. This paper evaluates vision-based TAS for continuous dance motion using video-derived skeleton sequences, eliminating the need for wearable sensors. A labeled dance dataset was created from video clips, including multi-person scenes where each performer was extracted and treated as an individual sequence. The dataset also incorporates duet performances, which introduce greater variation in execution and recording conditions.To examine how pose representation and temporal modeling affect segmentation quality, we evaluated two skeleton settings (BODY-25 and COCO-18) and two representative Temporal Action Segmentation (TAS) models [Multi-Stage Temporal Convolutional Network++ (MS-TCN++) and Frame-Action Cross-Attention Temporal Modeling (FACT)] under a unified protocol. Both models effectively segmented dance actions from skeleton data. FACT consistently achieved higher segment-level performance than MS-TCN++, particularly in F1 and Edit scores, while BODY-25 generally outperformed COCO-18 across both models. Using BODY-25, FACT achieved F1@10 and F1@25 scores of 94.6 and an F1@50 score of 90.6, compared to MS-TCN++'s scores of 89.7, 89.7, and 89.1, respectively. These results demonstrate that both pose representation and temporal modeling materially affect segmentation quality in continuous dance motion.
Abstract
This paper revisits the relationship between per capita income and technological-intensity structural transformation in manufacturing. It asks whether the income–cosphi relationship is better described by heterogeneous nonlinearity and smooth transition zones rather than by a common linear slope or hard income thresholds. Using a cross-country panel for 2014–2022, with a main sample of 482 observations from 65 countries and an expanded sample of 550 observations from 66 countries, the study employs two-way fixed effects as a benchmark, spline fixed-effects models as the main specification, threshold fixed-effects models as a heuristic tool, and a set of robustness checks including balanced versus unbalanced panels and Difference GMM. The results show that linearity is too restrictive. Nonlinear evidence is stronger in the main sample, the shape of the relationship differs between developed and developing economies, and threshold models do not support sharp breakpoints. Mechanism analysis further suggests that R&D conditions the nonlinear relationship more strongly than FDI. The paper contributes by reframing technological structural transformation as a nonlinear, heterogeneous, and conditional process rather than a universal income-threshold story.
Abstract
This study compares Filmbox (FBX), Alembic, and Pixar's Universal Scene Description (USD) as data-exchange mechanisms for visual effects (VFX) and digital content creation pipelines. Beyond their roles as file specifications, the analysis differentiates their production functions: FBX for editable rigged assets and animation, Alembic for baked time-sampled geometry, and USD for composed scene description using layers, references, variants, and sparse overrides. For reproducibility, a controlled benchmark was conducted in Blender. Identical geometry and animation were exported to each format, and software versions, scene complexity, repetition counts, and timing boundaries were explicitly reported. The benchmark evaluated two update scenarios—an animation revision and a layout/camera change—and two distinct operations: a full DCC import and a native USD layer reload. Across both scenarios, the sparse USD update layer constituted only a small fraction of the base file, while direct FBX and Alembic updates necessitated republishing nearly complete files. Full referenced-stage imports were comparable to or faster than legacy-format update imports, and reloading only a changed layer in an already-open USD stage was significantly more efficient than any full import. These results affirm USD's primary advantage in sparse composition and non-destructive updates. However, they do not support a universal claim that every USD load path is faster. FBX and Alembic remain effective for their specialized handoff roles, while USD is better suited for large, evolving scenes and parallel departmental workflows. Detailed file-size and timing measurements are reported in Section 4.
Abstract
This study investigates the educational potential of memes in the digital era and proposes effective strategies for integrating meme pedagogy into university general education. Using an integrative literature review, the study synthesizes research on memes, digital literacy, and digital native–digital immigrant theory to explore how meme-based learning can contribute to contemporary educational goals. Based on this synthesis, the study develops the Meme-Based Digital Literacy Pedagogy Framework. This framework illustrates the relationships among meme-based learning activities, digital literacy competencies, general education outcomes, and differentiated instructional strategies tailored for diverse learner groups. The findings suggest that meme-based activities—such as analysis, critique, creation, and collaborative projects—can foster critical comprehension, creative production, social collaboration, and digital citizenship. Furthermore, the framework emphasizes the necessity of differentiated instructional approaches for digital natives and digital immigrants to maximize educational effectiveness. The study concludes that memes can function as meaningful pedagogical resources, not merely as cultural artifacts, supporting digital literacy development and learner-centered general education. Implications for ethical implementation and future empirical validation are also discussed.
Abstract
This research introduces a novel convergent methodology integrating case-based observational analysis with traditional variable-based analysis. The study examines how lawful political contributions influence legislative decisions and voting behavior in the U.S. Senate, utilizing data on Political Action Committee (PAC) contributions and roll-call votes. First, correlation and regression analyses test whether PAC funding significantly associates with senators' voting patterns. Subsequently, a network analysis compares the network of PAC contributions among senators with their voting alignments based on legislative decisions. Findings indicate a positive association between PAC contributions and voting behavior: the amount of oil and gas PAC money a senator received is moderately correlated with the senator's pro-industry voting proportion (Pearson r = 0.48). At the network level, the similarity between the PAC contribution network and the voting network is slightly above random expectation but does not reach conventional statistical significance. Overall, the results suggest that PAC contributions associate with legislative voting at the individual level. The relational patterns among senators provide a complementary, albeit statistically inconclusive, perspective. This underscores the multi-layered nature of political influence in legislative processes and the value of examining contribution dynamics through senatorial interactions.
Abstract
This study investigated the relationship between Instagram usage motivation and psychological stress, specifically examining the mediating role of various Instagram posting behaviors. A survey was conducted by 'Survey Link,' a professional research company, targeting 300 adult men and women with Instagram posting experience. Data was collected from November 11 to November 13, 2024. Instagram usage motivations were categorized into four types: seeking interpersonal relationships, seeking information, seeking self-status, and seeking entertainment. Instagram postings were classified into three categories: self-pursuing postings, interactive ritual postings, and self-management postings. Results showed that information-seeking and self-status-seeking motivations were significantly associated with higher stress. Furthermore, among the different types of Instagram postings, only self-management postings significantly mediated the relationship between Instagram usage motivation and stress. These findings suggest that the impact of Instagram usage on mental health depends not only on usage time and posting behavior but, more fundamentally, on how distinct usage motivations shape these expressive behaviors.
Abstract
This study investigated the relationship between participation in classical music meditation and adults' quality of life (QoL), specifically examining whether poor sleep quality and chronic fatigue sequentially mediated this association. One hundred twenty adults who had experienced sleep or fatigue problems within the past six months participated in a six-week program, attending one 40-minute session per week. Sleep quality, chronic fatigue, and QoL were measured using the Pittsburgh Sleep Quality Index (PSQI), Fatigue Severity Scale (FSS), and WHOQOL-BREF, respectively. The reliability coefficients for these measures ranged from Cronbach's α = 0.81 to 0.91. Data were analyzed using SPSS 25.0 and PROCESS macro v.4.2 (Model 6) to test the sequential mediation model. Results indicated that participation in classical music meditation was significantly associated with a higher quality of life, mediated through improved sleep quality and reduced chronic fatigue. Specifically, the sequential pathway—classical music meditation participation → better sleep quality → lower chronic fatigue → higher quality of life —was supported. These findings suggest that music-based meditation can function as an integrative healing intervention, extending beyond a purely artistic experience. This has practical implications for education, counseling, and health promotion.
Abstract
With the advent of aging societies, fall-related injuries among older adults have emerged as a critical public health challenge. Conventional fall prevention education has predominantly relied on traditional, lecture-based approaches, which exhibit limited effectiveness in engaging older adult learners and facilitating knowledge retention. This study developed a card-based, gamified educational tool for fall prevention among older adults, grounded in Atkinson-Shiffrin's Multi-Store Model of Memory, Dual Coding Theory, and Cognitive Load Theory. To evaluate immediate educational efficacy, a between-subjects posttest-only comparative design was employed. An experimental group (A) receiving card-based gamified instruction was compared with a control group (B) receiving traditional lecture-based education. The study included 18 participants aged 65 years and older. Educational outcomes were assessed using a fall prevention knowledge test and the RIMMS-12 scale, an ARCS-based measure of Attention, Relevance, Confidence, and Satisfaction. Results demonstrated that the card-based gamified group achieved significantly higher total RIMMS-12 scores compared to the traditional lecture group (A: 51.44±3.50, B: 38.22±4.43, t=7.02, p<.001). Furthermore, the card-based group significantly outperformed the lecture group across all four RIMMS subscales (attention, relevance, confidence, and satisfaction) and demonstrated superior performance on the knowledge assessment (Welch's t=2.88, p=.0145, Hedges' g=1.29). These results indicate higher immediate posttest motivational and knowledge scores for the gamified approach, rather than definitive overall educational superiority. These findings suggest that interactive, game-based educational approaches may improve immediate learning outcomes among older adults. As this was a small-scale, posttest-only comparative experiment, the findings should be interpreted as preliminary evidence and verified in larger studies with follow-up assessments. This study is expected to contribute to the development of effective safety education content for older adult populations.
Abstract
This study examined the impact of two types of L1-mediated written corrective feedback—direct and indirect hybrid (indirect plus metalinguistic)—on L2 writing. Twenty-six Korean university students, at beginner to intermediate proficiency levels, completed a three-stage revision task: an initial draft, a first revision after receiving feedback, and a second revision without further assistance. Participants were divided into two groups, each receiving L1 feedback via ChatGPT: one group received direct feedback, and the other received indirect hybrid feedback. Uptake was analyzed across five writing components: content, organization, grammar, vocabulary, and mechanics. Both feedback types significantly improved writing after the first revision, with the most substantial uptake observed in mechanics and grammar. However, uptake decreased in the second revision, particularly in content and organization, and new or reintroduced errors were common. Notably, direct feedback proved more effective for beginner-level learners in later revisions, specifically for structural improvements and overall uptake. Conversely, intermediate learners benefited more from indirect hybrid feedback, especially in content development. These findings suggest that feedback effectiveness may vary with learner proficiency, offering pedagogical implications for customizing AI-mediated feedback in L2 writing instruction.
Abstract
Cardiac output (CO) is a crucial indicator for assessing hemodynamic status in intensive care units and clinical settings. However, conventional reference methods like echocardiography are limited by equipment accessibility and the challenges associated with repeated measurements. This study aimed to quantitatively analyze the relationships between physiological parameters derived from wearable electrocardiogram (ECG) devices and pulse wave-derived CO. The goal was to explore the feasibility of developing a wearable ECG-based CO estimation algorithm. The study included 30 healthy university students in their 20s, comprising both males and females, with no history of cardiovascular disease. The RR interval, QRS duration, and PR interval demonstrated statistically significant correlations with CO. Simple linear regression analysis revealed the following: a 1-ms increase in the RR interval was associated with an estimated decrease of 0.0072 L/min in CO; a 1-ms increase in the QRS duration was associated with an estimated increase of 0.0513 L/min; and a 1-ms increase in the PR interval was associated with an estimated decrease of 0.0346 L/min in CO. Our findings confirm that physiological parameters extracted from wearable ECG devices exhibit significant linear relationships with pulse wave-based CO. This supports the technical feasibility of noninvasive CO estimation using wearable ECG signals.
Abstract
Traditional manpower-based approaches to apple disease control face growing challenges due to the increased frequency of anomalous weather events, attributable to climate change, and labor shortages caused by aging rural populations. Specifically, the prevalence of high-temperature and high-humidity environments has led to a surge in domestic apple pests and diseases, resulting in significant economic losses and a sharp decline in total apple production. This paper proposes an AI-based prediction model that utilizes agro-meteorological data and a specialized dataset, focused on the "15-day window before and after the first precipitation," for the proactive forecasting of three major apple diseases: Marssonina blotch, White rot, and Anthracnose. The proposed method consistently demonstrated superior performance across all target diseases in terms of prediction accuracy, sensitivity, and specificity. Notably, the model achieved a sensitivity of 0.9750 for Anthracnose and a specificity of 0.9355 for White rot, validating its potential for practical on-farm application. This study is expected to advance eco-friendly precision agriculture by enabling farmers to optimize pest control timing and reduce unnecessary pesticide usage through the proactive prediction of disease outbreaks based on meteorological variations.