GIST Scholar
Not a member yet
30271 research outputs found
Sort by
Aerosol-driven North Pacific High anomaly enhances sea ice loss in the Chukchi Sea
No full textThe Arctic has warmed significantly faster than the rest of the globe, leading to rapid sea ice decline. Anthropogenic aerosols are traditionally viewed as cooling agents that do not contribute to Arctic sea ice loss. Here we investigate how aerosol-induced changes in atmospheric circulation patterns contribute to Arctic sea ice decline using a fully-coupled global earth system model. We compared single forcing experiments to examine individual and combined effects of greenhouse gases and anthropogenic aerosols. Aerosols contribute to intensification of the North Pacific anticyclone, which enhances heat transport into the Arctic through the Bering Strait. When combined with greenhouse gas-induced warming, aerosols have a greater impact on Arctic sea ice decline in the western Chukchi Sea compared to when either forcing acts independently. This compound effect challenges the traditional view of aerosols as solely cooling agents, demonstrating that anthropogenic aerosols can accelerate Arctic sea ice melting.TRUEsciescopu
Design and Analysis of a Low-Jitter High-Gain Latch-Based Time Amplifier
No full textThis paper presents design and analysis of a latch-based time amplifier (TA). While prior studies have primarily focused on TA gain and output characteristics, this work extends the analysis to include input-referred jitter, input dynamic range, latency, and gain behavior. Analytical equations are derived to facilitate comparisons with other TA architectures, offering insights into the performance trade-off of latch-based TA. The proposed models are validated through both simulations and measurements. A prototype latch-based TA is fabricated in a 28-nm CMOS LP process to demonstrate the effectiveness of the analysis.FALSEsciescopu
Short-Segment Speaker Verification via Multi-Perspective Feature Fusion from Self-Supervised Speech Representations
No full textSpeaker verification systems have shown impressive performance on long utterances. However, their performance significantly degrades on short-duration utterances. To address this issue, a multi-resolution encoder has been proposed to extract low-level representations at multiple temporal resolutions and condition the hidden representations of ECAPA-TDNN, achieving state-of-the-art performance for utterances shorter than 2 seconds. In this work, we propose a multi-perspective feature fusion method to enhance the baseline system. The method extracts fused features by multiplying a learnable matrix with representations from a pre-trained self-supervised model and injects them as conditional information into each SE-Res2Block of ECAPA-TDNN. Experimental results on VoxCeleb1-O demonstrate that our method further improves performance on short utterances compared to the baseline.Master1. Introduction 1
1.1 Introduction 1
2. Baseline 4
2.1 ECAPA-TDNN with Multi-Resolution Encoder 4
2.1.1 Multi-resolution encoder 5
2.1.2 Adapter module 6
2.2 WavLM-Base+ 7
3. Proposed Method 9
3.1 Multi-perspective feature fusion 9
3.2 Integrating conditional features 11
4. Experiments 13
4.1 Dataset 13
4.2 Training strategy 13
4.3 Implementation details 15
4.4 Evaluation protocol 15
4.4.1 Short-segment test utterance generation 15
4.4.2 Evaluation metric 17
5. Results and ablation studies 18
5.1 Experimental Results 18
5.2 Abalation Study 19
6. Conclusion 23
References 2
Dental pulp stem cell secretome inhibits mitophagy-induced hippocampal neural injury during hypoxia
No full textHypoxic stress induces neuronal damage by increasing mitochondrial reactive oxygen species (mtROS), triggering mitophagy-associated cell death, and promoting neuroinflammation. However, the neuroprotective potential of human dental pulp stem cell (hDPSC) secretome in this process remains unclear. Here, we show that the hDPSC secretome mitigates hypoxic stress-induced neuronal injury by modulating mitophagy and inflammatory pathways. Proteomic analysis revealed key therapeutic proteins in the hDPSC secretome, which appear to reduce mtROS levels and suppress mitophagy markers, including PHB2, BCL2L13, BNIP3, and LC3-II. CoCl2-induced activation of the TLR4-NF-kappa B pathway increased pro-inflammatory cytokines, promoting cell death, but the hDPSC secretome inhibited this activation while enhancing anti-apoptotic proteins to support neuronal survival. Furthermore, the hDPSC secretome restored neuronal and synaptic markers, including neurofilament heavy chain, synaptophysin, and PSD95, contributing to neuronal recovery and synaptic integrity. This study provides evidence that the hDPSC secretome alleviates hypoxic stress-induced neuronal damage through a multifaceted neuroprotective mechanism. These findings indicate its potential as a therapeutic strategy for neurodegenerative conditions.FALSEsciescopu
First-principles, data-guided screening and catalyst design: Unveiling energetic trade-offs in ammonia decomposition
No full textAmmonia decomposition is a promising route for CO2-free hydrogen production, but the development of efficient and cost-effective catalysts remains a challenge. Here, we employed a dual approach combining computational screening and free energy analysis to identify optimal catalysts, which were then validated experimentally. Pearson correlation coefficient analysis revealed a volcano-type relationship between NH3 dissociation energy (Ediss,NH3) and N2 adsorption energy (Ead,N2), highlighting Ru as the most effective monometallic catalyst. Reactor tests using supported Ru, Rh, Ir, Ni, and Co catalysts confirmed these predictions, with Ru exhibiting the highest NH3 conversion and a strong correlation between turnover frequency, activation energy (Ea), and electronic descriptors. Using this validated approach, we extended our analysis to bimetallic systems, identifying Ru-Ni alloys, as a promising alternative with balanced NH3 activation and N2 desorption. These findings demonstrate the effectiveness of combining computational and experimental methods to design high-performance NH3 decomposition catalysts. Further refinement of Ru-Ni alloy synthesis and structural control could enhance catalytic activity, supporting scalable hydrogen production. © 2025 Elsevier Inc.FALSEsciescopu
Convenient gearbox fault diagnosis under random variable speeds: A motor current nonlinear harmonic approach
No full textMotor current signature analysis (MCSA) techniques are gradually utilized for gearbox fault detection, given the convenience of current clamp installation and the clarity of signals. However, applying the MCSA to gearbox fault diagnosis presents substantial challenges. Specifically, fault features can be overwhelmed by both the fundamental frequency of the current and complex sidebands, and are susceptible to variations under variable speed conditions. To address these issues, an analytical model of current signals for localized gear faults is established, and a time–frequency analysis method is proposed for diagnosing gear faults. First, considering both amplitude modulation and frequency modulation effects, an analytical model of d-axis current signals under fault is established, which helps mitigate the influence of current fundamental frequency and complex sidebands. Second, a gear fault detection method called Iterative Vold-Kalman Filter is proposed, which combines the surrogate test with the Vold-Kalman Filter to solve the problem of fault representation under variable speed conditions. Finally, the proposed method is verified by simulated and experimental data from gearbox fault cases, and compared with classical algorithms to highlight the superiority of the proposed method. Overall, the proposed method enables quick and accurate detection of gear faults under variable speed conditions. © 2024 Elsevier LtdFALSEsciescopu
TelePulse: Enhancing the Teleoperation Experience through Biomechanical Simulation-Based Electrical Muscle Stimulation in Virtual Reality
No full textThis paper introduces TelePulse, a system integrating biomechanical simulation with electrical muscle stimulation (EMS) to provide precise haptic feedback for robot teleoperation tasks in virtual reality (VR). TelePulse has two components: a physical simulation part that calculates joint torques based on real-time force data from remote manipulators, and an electrical stimulation part that converts these torques into muscle stimulation. Two experiments were conducted to evaluate the system. The first experiment assessed the accuracy of EMS generated through biomechanical simulations by comparing it with electromyography (EMG) data during force-directed tasks, while the second experiment evaluated the impact of TelePulse on teleoperation performance during sanding and drilling tasks. The results suggest that TelePulse provided more accurate stimulation across all arm muscles, thereby enhancing task performance and user experience in the teleoperation environment. In this paper, we discuss the effect of TelePulse on teleoperation, its limitations, and areas for future improvement. © 2025 Copyright held by the owner/author(s)
Bioaccumulation of trace elements in the Mekong river floodplain, Laos: implications for human health
No full textTo investigate trace element bioaccumulation in floodplain regions and its impact on human health, samples of groundwater, soil, and food crops were taken from two floodplain locations (Champasak (CH) and Attapeu (AT)) along the Mekong River in the Lao People's Democratic Republic (Lao PDR). Inductively coupled plasma–mass spectrometry (ICP–MS) analyzed trace element concentrations in groundwater and food crops, while soil samples were examined using ICP–optical emission spectrometry (ICP–OES). The concentrations of arsenic (As), iron (Fe), and manganese (Mn) in groundwater surpassed WHO guidelines by 13.3%, 5%, and 56.6%, respectively, in CH (n = 60) and by 23.9%, 1.1%, and 38%, respectively, in AT (n = 92). In soil, 67%, 33.3%, and 11.1% of samples from CH (n = 9) and 22.6%, 9.6%, and 3.2% from AT (n = 31) exceeded permissible levels for As, copper (Cu), and chromium (Cr). Geographical variations contributed to discrepancies in environmental quality between the two areas, as demonstrated by pollution index evaluations. The study region is at high ecological risk, with both children and adults potentially exposed to non-carcinogenic and carcinogenic risks via the food chain, especially through ingestion. These findings are crucial for informing policy and enhancing risk management strategies in the floodplain regions. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2025.FALSEsciescopu
Sub-1-volt, optically readable active Tamm plasmon resonators for free-space multispectral image processing
No full text고급 시각 데이터 처리에 대한 수요가 증가함에 따라, 전력 소비, 적응성, 대용량 시각 데이터 처리 능력 측면에서 새로운 컴퓨팅 시스템의 필요성이 대두되고 있다. 광 컴퓨팅은 고대역폭 광통신과 국소적 정보 처리를 통합함으로써, 처리 속도와 에너지 효 율 측면에서 상당한 향상을 제공한다. 그러나 이러한 진보에도 불구하고, 멀티스펙트럼 데이터에서의 노이즈 제거, 스펙트럼 필터링, 대비 향상과 같은 과제를 해결하기 위한 추가적인 기술 개발이 요구된다. 본 연구에서는 1V 이하의 구동 전압으로 작동하는 능 동형 Tamm 플라스몬 공명기를 활용하여 높은 적응성을 갖춘 멀티스펙트럼 영상 처리 시스템을 개발하는 데 중점을 두었다. 이 구조는 이론적으로 99프로에 달하는 높은 변 조 깊이와 PEDOT:PSS의 장기적이고 비휘발성 메모리 특성을 결합함으로써, 연속적인 전기펄스입력하에서도 256단계의시냅스상태를안정적으로구현할수있음을확인하 였다. 이러한 제어 가능성은 노이즈 감소뿐만 아니라 고성능 스펙트럼 필터링 및 시각적 대비 향상에도 기여함으로써, 광 컴퓨팅 기술에서의 실용성 및 성능을 향상시킨다. 제안 된 고적응형 멀티스펙트럼 영상 처리 시스템은 복잡한 영상 처리 작업에 특화되어 보다 효율적이고 확장이 가능한 반응성 높은 차세대 광기반 기술로서의 가능성을 보여준다.|The growing requirement for advanced visual data processing highlights the need for computing systems in terms of power consumption, adaptability, and handling large volumes of visual data. Optical computing integrates high bandwidth optical communi- cation with localized information processing, offering substantial improvements in pro- cessing speed and energy efficiency. Despite these advancements, further development is needed to address difficulties related to denoising, spectral filtering, and contrast improvement in multispectral data. Here, we focus on developing a highly adaptable multispectral image processing by utilizing active Tamm plasmon resonators operating under sub-1 volt conditions. By combining the high modulation depth, theoretically reaching 99%, with the long-term and non-volatile memory properties of PEDOT:PSS, the stability of 256 synaptic states is demonstrated under continuous electrical pulse inputs. This tunability not only improves noise reduction but also facilitates high- performance spectral filtering and visual contrast in optical computing technologies. A highly adaptable multispectral image processing system shows promise for more effi- cient, scalable, and responsive technologies tailored to the complex image processing task.Master1 Introduction 1
1.1 Optical computing technologies compared with traditional electronic computing architectures 1
1.2 Adaptive photonic systems for multispectral image processing 2
1.2.1 Optical neuromorphic computing based on active Tamm plas- mons 2
1.2.2 High-Q Tamm plasmon resonators for spectral filtering 3
2 Optimization of Active Tamm plasmon resonator for multispectal op- eration 7
2.1 Structural configuration and operating principle 7
2.1.1 Optically tunable modulator with PEDOT:PSS 7
2.1.2 Redox-induced doping of PEDOT:PSS 8
2.2 Multispectral tuning via last layer thickness modulation 10
3 Optical Synaptic Characteristics for denoising Functions 15
3.1 Optical Synaptic Characteristics for denoising Functions 15
3.2 Optical synaptic dynamics: EPSP/IPSP and paired-pulse facilitation . 16
3.3 Synaptic plasticity and memory stability 17
4 Multispectral Image denoising system with Synaptic Modulation 22
4.1 Selectively multispectral image processing using MOSR filter 22
4.2 Fabrication of MOSR filter 25
4.2.1 Optical simulation 25
4.2.2 Optical and Electrochemical measurements 26
4.2.3 Optical characterization and measurement 26
4.2.4 Neural network simulation and implementation 27
– v –
5 Conclusion 29
References 30
Acknowledgements 3
New particle formation prediction by using a machine learning method
No full textNew Particle Formation (NPF) is a major atmospheric process that impacts the climate, atmospheric chemistry, and human health. NPF events can occur anywhere around the world, and can make up to 50% of all aerosol number concentration. NPF events are considered a strong precursor for cloud condensation nuclei (CCN), which in turn affect air pollution and cloud formation. This study develops complex models to predict NPF events in the urban Gwangju site using meteorological and gas species information. Classified NPF days datasets from Gwangju, South Korea (2014-2025), along with temperature, pressure, relative humidity, solar radiation, cloud coverage, and various gas species concentrations, were implemented in machine learning (ML) models. Data augmentation was achieved with SMOTE, while feature selection was conducted by Kruskal-Wallis and Mutual Information (MI) tests. Results show that the Extreme Gradient Boosting (XGB) model illustrated the highest classification accuracy (81.82%) for the urban Gwangju site. The regression analysis performed on the models illustrated satisfactory results, with a moderate coefficient of determination (R2) score of 0.58 for the test dataset. Overall, basic features such as solar radiation, cloud coverage, temperature, heat index, wind speed, CO, and PM2.5 are observed to be effective, minimal predictors for the NPF events.Master1. INTRODUCTION 6
2. MATERIALS AND METHODS . 10
2.1. Site description 11
2.2. Data processing 13
2.2.1. Data classification 13
2.2.2. SMOTE 15
2.2.3. Feature selection . 16
2.2.5. Hyperparameter tuning 19
2.2.6. Machine Learning models 20
3. RESULTS AND DISCUSSION . 24
3.1. 2024-2025 Gwangju NPF investigation . 24
3.2. Classification. 28
3.3. Regression 32
4. CONCLUSIONS 37
REFERENCES . 3