GIST Scholar
Not a member yet
30271 research outputs found
Sort by
FlowSE: Flow Matching-based Speech Enhancement
No full textDiffusion probabilistic models have shown impressive performance for speech enhancement, but they typically require 25 to 60 function evaluations in the inference phase, resulting in heavy computational complexity. Recently, a fine-tuning method was proposed to correct the reverse process, which significantly lowered the number of function evaluations (NFE). Flow matching is a method to train continuous normalizing flows which model probability paths from known distributions to unknown distributions including those described by diffusion processes. In this paper, we propose a speech enhancement based on conditional flow matching. The proposed method achieved the performance comparable to those for the diffusion-based speech enhancement with the NFE of 60 when the NFE was 5, and showed similar performance with the diffusion model correcting the reverse process at the same NFE from 1 to 5 without additional fine tuning procedure. We also have shown that the corresponding diffusion model derived from the conditional probability path with a modified optimal transport conditional vector field demonstrated similar performances with the NFE of 5 without any fine-tuning procedure. © 2025 IEEE
직접 암모니아 연료전지의 Perovskite 구조 기반 Anode의 B-site 화학양론 제어를 통한 성능 개선 연구
No full textAmmonia has high energy density and the advantage of utilizing existing infrastructure, making it increasingly significant as a direct fuel for energy production in solid oxide fuel cell (SOFC). However, ammonia-fueled SOFC have a lower power density compared to hydrogen fuel cells. To address this issue, this study investigates Sr2Fe1.6M0.5O6-δ, an electrode material based on Sr2Fe1.5M0.5O6-δ with high conductivity and stability under reducing environments, by further increasing the Fe content. The structural stability and the formation of Fe nanoparticles on the electrode surface were confirmed through XRD, SEM, and TEM analyses. The electrochemical performance and catalytic properties of the fuel cells were evaluated. As the Fe content increased, more Fe nanoparticles were formed under reducing conditions, enhancing both electrochemical and catalytic performance. Notably, the electrochemical performance of the DA-SOFC increased by approximately 2.4 times, achieving performance comparable to fuel-electrode-supported DA-SOFCs. In conclusion, this study demonstrates a viable solution to overcome the performance limitations of oxide electrode as anode.Master초 록 ․․․․․․․․․․․․․․․․․․․․ i
목 차 ․․․․․․․․․․․․․․․․․․․․ ii
그림목차 ․․․․․․․․․․․․․․․․․․․․ iv
표 목 차 ․․․․․․․․․․․․․․․․․․․․ vi
I. 서 론 ․․․․․․․․․․․․․․․․․․․․ 1
1. 1. 연구 배경 ․․․․․․․․․․․․․․․․․․․․ 1
1. 1. 1. 신ㆍ재생에너지 ․․․․․․․․․․․․․․․․․․․․ 1
1. 1. 2. 에너지 저장 기술 ․․․․․․․․․․․․․․․․․․․․ 5
1. 1. 3. 연료전지 ․․․․․․․․․․․․․․․․․․․․ 21
1. 1. 4. 고체 산화물 연료전지 ․․․․․․․․․․․․․․․․․․․․ 26
1. 2. 연구 목적 ․․․․․․․․․․․․․․․․․․․․ 31
1. 2. 1. 직접 암모니아 고체 산화물 연료전지 ․․․․․․․․․․․․․․․․․․․․ 31
1. 2. 2. 직접 암모니아 고체 산화물 연료전지 연료극 ․․․․․․․․․․․․․․․․․․․․ 38
II. 직접 암모니아 산화물 연료전지 제작 ․․․․․․․․․․․․․․․․․․․․ 44
2. 1. 연료극 물질 합성 및 분석 ․․․․․․․․․․․․․․․․․․․․ 44
2. 2. 고체 산화물 연료전지 제작 ․․․․․․․․․․․․․․․․․․․․ 48
III. 직접 암모니아 산화물 연료전지 연료극 분석 ․․․․․․․․․․․․․․․․․․․․ 51
3. 1. 연료극 물리적 특성 분석 ․․․․․․․․․․․․․․․․․․․․ 51
3. 2. 연료극 전기화학적 분석 및 성능 평가 ․․․․․․․․․․․․․․․․․․․․ 56
3. 3. 연료극 촉매 성능 분석 ․․․․․․․․․․․․․․․․․․․․ 64
IV. 결론 ․․․․․․․․․․․․․․․․․․․․ 67
V. 참고 문헌 ․․․․․․․․․․․․․․․․․․․․ 6
Multiplex Optical Unclonable Functions: Advances and Perspectives in Optics and Photonics for Hardware Security
No full textWith the accelerating expansion of connectivity, the need for advanced cyber-physical security technologies that bridge the digital and physical worlds is becoming more crucial than ever. Physically unclonable functions (PUFs) leveraging nanotechnologies and photonic technologies are emerging as practical and deployable hardware security solutions that go beyond software-based hardware security. Optics- and photonics-based PUFs (often referred to as optical PUFs) offer a range of characteristics beneficial to multiplex strategies that incorporate multilevel and multimodal approaches, based on their fundamental optical and photonic properties. PUFs are categorized as multilevel when multiple responses are generated from a single type of challenge, such as a light source or optical tool. In contrast, they are classified as multimodal when multiple responses are created from various types of challenges involving a combination of different light sources and optical tools. This review highlights recent advances and progress in integrating emerging materials, detection tools, authentication methods, imaging techniques, reconfigurability, and unclonability, which are essential for advancing next-generation multiplex PUFs. To facilitate real-world applications and to support deployment in real-world senarios, we also discuss existing limitations in multiplex optical PUFs and explore potential strategies to overcome such challenges, supporting our increasingly hyperconnected society.FALSEsciescopu
Investigation of Mineralogical Properties and Arsenic Speciation in Soils from an Abandoned Gold Mine Using Synchrotron-Based X-ray Techniques
No full text본 연구는 한국에 위치한 한 폐금속광산의 비소오염 토양을 대상으로, 방사광가속기를 이용한 X-선 회절(XRD), X-선 흡수 분광법(XAS), 편광현미경 및 화학성분 분석을 통합적으로 활용하여 토양의 광물학적·화학적 특성과 비소의 존재형태를 규명하였다. 정량 X-선 회절(QXRD) 분석 결과, 토양은 입도에 따라 석영과 운모를 주 구성광물로 하고, 활석, 백운석, 녹니석 등이 다양하게 포함되어 있는 것으로 나타났다. 이러한 QXRD 결과는 편광현미경 분석을 통해 해당 광물들의 존재가 시각적으로 확인되었고, XRF 분석을 통해 광물 조성과 일치하는 원소 성분이 검출되어 잘 뒷받침되었다. XAS 분석 결과, 비소는 주로 스코로다이트(FeAsO4·2H2O) 광물 형태로 존재하며, 일부 유비철석(FeAsS) 광물 및 비산염이온(AsO43-) 흡착 형태가 혼재하고 있는 것으로 확인되었다. 75 μm 미만의 세립질 시료에서는 비소가 최대 5,000 mg/kg 이상 고농도로 농집되어 있어, 비산 및 강수에 의한 유실 위험성이 제기되며, 약알칼리 조건(pH 8.1)에서는 비산염의 용출이 활발히 일어남에 따라 자연적인 저감 효과가 제한적일 수 있음이 확인되었다. 본 연구는 이러한 결과를 바탕으로 효과적인 환경대응 전략의 필요성을 제안하였으며, 방사광 기반 X-선 분석기법이 지질환경 분야에서 미량 오염물질의 존재형태 규명에 매우 유용한 도구임을 강조하였다.TRUEscopuskc
A Dual-Attention RNN for Repeat PCI Prediction Using EHRs
No full textPercutaneous coronary intervention (PCI) is a standard treatment for significant coronary artery disease (CAD); however, predicting the need for repeat PCI remains clinically challenging. Traditional survival models, such as the Cox proportional hazards model, rely on static baseline features and fail to capture the dynamic nature of patient trajectories following initial PCI. In this study, we propose DA-RNN-Surv, a novel interpretable survival analysis framework leveraging dual attention recurrent neural networks to incorporate longitudinal electronic health record (EHR) data. Using a real-world cohort of 6,252 PCI patients collected over a 10-year period, DA-RNN-Surv achieved significantly higher predictive performance, with concordance indices ranging from 0.722 at month 3 to 0.894 at month 24, outperforming traditional Cox (0.525 to 0.635) and DeepSurv (0.506 to 0.616) models. Importantly, DA-RNN-Surv provides interpretability through visit-level and feature-level attention weights, identifying key clinical timepoints (e.g., baseline and follow-up months) and influential variables such as medication adherence and comorbidities. These findings demonstrate the clinical value of explicitly modeling temporal information, highlighting DA-RNN-Surv’s potential for enhancing personalized patient management and preventive intervention strategies post-PCI. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026
Over 60% precipitation transformed into terrestrial water storage in global river basins from 2002 to 2021
No full textThe crucial role of precipitation as a primary driver for terrestrial water cycle is well-established. However, quantifying the transformation of daily precipitation into terrestrial water storage remains a challenge. Here we address this by introducing a quantitative metric, average daily fraction of precipitation transformed into terrestrial water storage, providing an important advancement into the dynamics of water storage by utilizing the enhanced terrestrial water storage statistical reconstruction method and water storage data from the Gravity Recovery and Climate Experiment satellites and their follow-on mission. This study reveals that approximately 64% of land precipitation contributes to terrestrial water storage in global 121 river basins from 2002 to 2021, with evident variations observed across different climatic and geographical regions. Our findings deepen perception into the complex interactions between precipitation, land surface processes, and climate change, offering valuable implications for future water resource management and hydrological modeling.TRUEsciescopu
Effects of attention-demanding task-based technique on reorientation in virtual reality
No full textReorientation techniques in virtual reality (VR) enable users to explore virtual spaces that are larger than the physical spaces within limited physical spaces. Increasing the rotation gain, which alters the user's orientation between physical and virtual spaces, enables users to navigate a larger virtual space. However, as a significant rotation gain may disrupt the user's sense of presence and cause discomfort, it is used at limited values. To address this issue, we propose an attention-demanding task-based reorientation technique that leverages the human perceptual phenomenon of inattentional blindness. We design attention-demanding tasks that require users to focus on specific types of attention (visual, mental, and physical) and verify that tasks are effective in demanding the intended attention type. In addition, we measure the user noticeability, presence, and VR sickness. The results show that users are less likely to notice reorientation during attention-demanding tasks than during tasks without specific attentional demands, which also avoids undesired effects such as presence degradation and increased VR sickness. Our findings suggest VR content design guidelines for seamlessly integrating reorientation techniques into VR interactive scenarios, thereby increasing the opportunities for reorienting users in virtual environments.FALSEsciessciscopu
Submicron Patterning Techniques for Perovskite Optoelectronics From Materials Challenges to Scalable Device Integration
No full textHalide perovskites are promising materials for future optoelectronic and display systems. They combine strong optical absorption, mechanical flexibility, low-cost solution processing, and tunable band structures. To serve emerging applications such as high-resolution AR/VR micro displays, optical logic gates, and neuromorphic vision processors, patterning methods should deliver sub-100nm resolution, more than 99% photoluminescence retention in multilayer stacks, and alignment precision better than 500nm. Sensitivity to polar solvents, heat, and chemical treatments limits compatibility with conventional photolithography and often leads to degradation during fabrication. This review surveys recent patterning strategies that address these issues. The discussion centers on four criteria: solvent compatibility, resolution, alignment accuracy, and process methodology. Mask-based approaches such as direct photolithography and mold-guided growth are compared with mask-free routes, including inkjet printing and laser-induced crystallization. Emerging directions include perovskite-compatible photoresists and hybrid schemes that couple nanoscale precision with wafer-scale scalability, outlining a path toward high-density, multifunctional, and vision-adaptive optoelectronic platforms. © 2025 The Author(s). Advanced Optical Materials published by Wiley-VCH GmbH.TRUEsciescopu
Solution-State Programming of Crystallinity in Conjugated Polymer Nanowires for Customized Electronics
No full textRecent advances in solution-state electron microscopy have enabled direct visualization of molecular assembly of polymers. However, controlling the crystallinity within the polymer nanostructures which result in electronic and mechanical properties remains challenging. Here, we demonstrate a strategy to control the degree of crystallinity in conjugated polymer nanowires within solution, allowing selective formation of highly ordered (≈70%) and partially ordered (≈20%) assemblies. These nanowires were visualized by transmission electron microscopy (TEM) and analyzed in pixel scale, revealing distinct molecular arrangements that influence π-π stacking coherence and interchain alignment. When incorporated into thin-film transistors, the highly crystalline nanowires exhibited enhanced charge mobility, while the less ordered ones retained high stretchability. Furthermore, their controlled blending enabled simultaneous optimization of electrical performance and mechanical flexibility. This microscopy-guided approach highlights the potential of solution-phase crystallinity programming as a tool for designing multifunctional polymer assemblies with tunable structure-property relationships
Novelty Detection in Underwater Acoustic Environments Using Out-of-Distribution Detector for Neural Networks
No full text본 연구에서는 수중 음향 환경에서 학습되지 않은미지의 음향 신호를 효과적으로 탐지하기 위한 ODIN 기반의 이상 음원 탐지 프레임워크를 제안한다. 구체적으로, 사전 학습된 분류기의 소프트맥스 출력에 온도 스케일링과 입력 교란을 적용하여 훈련된 클래스와 미지의 클래스간의 출력 차이를 유도하고, 최대소프트맥스 확률을 이상 점수로 활용하여 이상 탐지를 수행한다.FALSEscopuskc