• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.9-12
• /
• 2023

최근 영상 콘텐츠를 통해 영상물의 메시지뿐 아니라 메시지의 형식을 통해 전달된 감정이 시청하는 사람의 심리 상태에 영향을 주고 있다. 이에 따라, 영상 콘텐츠의 감정을 분류하는 연구가 활발히 진행되고 있고 본 논문에서는 대중적인 영상 스트리밍 플랫폼 중 하나인 유튜브 영상을 7가지의 감정 카테고리로 분류하는 여러 개의 영상 데이터 중 각 영상 데이터에서 오디오와 이미지 데이터를 각각 추출하여 학습에 이용하는 멀티 모달 방식 기반의 영상 감정 분류 모델을 제안한다. 사전 학습된 VGG(Visual Geometry Group)모델과 ViT(Vision Transformer) 모델을 오디오 분류 모델과 이미지 분류 모델에 이용하여 학습하고 본 논문에서 제안하는 병합 방법을 이용하여 병합 후 비교하였다. 본 논문에서는 기존 영상 데이터 감정 분류 방식과 다르게 영상 속에서 화자를 인식하지 않고 감정을 분류하여 최고 48%의 정확도를 얻었다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2024.01a
• /
• pp.45-46
• /
• 2024

본 논문에서는 Self-Attention 기반 딥러닝 기법인 Dense Prediction Transformer(DPT) 모델을 MVTec Anomaly Detection(MVTec AD) 데이터셋에 적용하여 실제 산업 제품 이미지 내 이상 부분을 분할하는 연구를 진행하였다. DPT 모델의 적용을 통해 기존 Convolutional Neural Network(CNN) 기반 이상 탐지기법의 한계점인 지역적 Feature 추출 및 고정된 수용영역으로 인한 문제를 개선하였으며, 실제 산업 제품 데이터에서의 이상 분할 시 기존 주력 기법인 U-Net의 구조를 적용한 최고 성능의 모델보다 1.14%만큼의 성능 향상을 보임에 따라 Self-Attention 기반 딥러닝 기법의 적용이 산업 제품 이상 분할에 효과적임을 입증하였다.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.82-95
• /
• 2024

As automated essay scoring (AES) has progressed from handcrafted techniques to deep learning, holistic scoring capabilities have merged. However, specific trait assessment remains a challenge because of the limited depth of earlier methods in modeling dual assessments for holistic and multi-trait tasks. To overcome this challenge, we explore providing comprehensive feedback while modeling the interconnections between holistic and trait representations. We introduce the DualBERT-Trans-CNN model, which combines transformer-based representations with a novel dual-scale bidirectional encoder representations from transformers (BERT) encoding approach at the document-level. By explicitly leveraging multi-trait representations in a multi-task learning (MTL) framework, our DualBERT-Trans-CNN emphasizes the interrelation between holistic and trait-based score predictions, aiming for improved accuracy. For validation, we conducted extensive tests on the ASAP++ and TOEFL11 datasets. Against models of the same MTL setting, ours showed a 2.0% increase in its holistic score. Additionally, compared with single-task learning (STL) models, ours demonstrated a 3.6% enhancement in average multi-trait performance on the ASAP++ dataset.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.24 no.3
• /
• pp.33-38
• /
• 2010

In this paper, the nature of iron loss in electrical steel during alternating field excitation is investigated more precisely. The exact definition of AC iron loss is cleared by accurately measuring the iron loss for conditions of both the sinusoidal magnetic field and sinusoidal magnetic flux density. The results of this approach to iron loss calculations in electrical steel are compared to experimentally-measured losses. In addition, an inverse hysteresis model considering eddy current loss was developed to analyze the iron loss when the input is the voltage source. With this model, the inrush current in the inductor or transformer as well as the iron loss can be calculated.

• KIEE International Transactions on Electrophysics and Applications
• /
• v.5C no.1
• /
• pp.1-7
• /
• 2005

This paper describes the lightning surge analysis model of extra high voltage GIS using EMTP. Various lightning current parameters were investigated in order to confirm the impact on the lightning surge analysis such as lightning current amplitude, waveform, size of GIS, tower footing resistance and surge arresters. The multi-story tower model and EMTP/TACS model were introduced for the simulation of dynamic arc characteristics. The margin between the maximum overvoltage and BIL of the GIS was about 10 percent and the margin between the maximum overvoltage and BIL of the transformer was 21 percent.

• Proceedings of the Korean Society Of Semiconductor Equipment Technology
• /
• 2003.12a
• /
• pp.182-185
• /
• 2003

We have suggested an equivalent circuit model for switching performance of bipolar spin transistor composed of a nonmagnetic metal film (N) sandwiched between two ferromagnetic metal films (F1 and F2). The 'ON' or 'OFF' operation of this equivalent circuit model is simulated by depending on the orientation of the magnetization of F1 and F2 rather than the strength of the external magnetic filed. Changing the coupling coefficient, turn number of two inductances, (L1:L2) like a transformer, and parallel variable resistance R4 connected to L2 at the collector region, we can explain the magnetic characteristics and the dependence of magneto resistance ratio on the orientation of spin-polarized electrons.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.26 no.5
• /
• pp.91-96
• /
• 2012

The surface of bushing is contaminated with rain, dust, salt and others. A bushing with contaminations in air is serious problem in insulation. Therefore, it is important to understand the inspection and diagnoses of the safety. The ultra-violet rays(UV) camera has attracted interest from the view point of easy judgement. In this paper, we will report on the corona discharge characteristics of surface flashover model with contaminations in air. Also, UV images of discharge and corona pulse count in air are analyzed using prototype UV camera of Korea and a UV sensor with an optic lens. These results are studied at both AC and DC voltage under a non-uniform field.

• Journal of Power Electronics
• /
• v.9 no.1
• /
• pp.85-92
• /
• 2009

This paper studies the method to predict voltage variation that can be presented in the case of operating a small-sized wind turbine in grid connection to the isolated small-sized power system. In order to do this, it makes up the simplified simulation model of the existing power plant connected to the isolated system, load, transformer, and wind turbine on the basis of PSCAD/EMTDC and compares them with the operating characteristics of the actual established wind turbine. In particular, it suggests a simplified model formed with equivalent impedance of the power system network including the load to analytically predict voltage variation at the connected point. It also confirms that the voltage variation amount calculated by the suggested method accords well with both simulation and actually measured data. The results can be utilized as a tool to ensure security and reliability in the stage of system design and preliminary investigation of a small-sized grid connected wind turbine.

• ETRI Journal
• /
• v.43 no.6
• /
• pp.1038-1048
• /
• 2021

We propose an end-to-end neural coreference resolution for the Korean language that uses an attention mechanism to point to the same entity. Because Korean is a head-final language, we focused on a method that uses a pointer network based on the head. The key idea is to consider all nouns in the document as candidates based on the head-final characteristics of the Korean language and learn distributions over the referenced entity positions for each noun. Given the recent success of applications using bidirectional encoder representation from transformer (BERT) in natural language-processing tasks, we employed BERT in the proposed model to create word representations based on contextual information. The experimental results indicated that the proposed model achieved state-of-the-art performance in Korean language coreference resolution.

• ETRI Journal
• /
• v.46 no.1
• /
• pp.96-105
• /
• 2024

We propose a method to automatically predict Alzheimer's disease from speech data using the ChatGPT large language model. Alzheimer's disease patients often exhibit distinctive characteristics when describing images, such as difficulties in recalling words, grammar errors, repetitive language, and incoherent narratives. For prediction, we initially employ a speech recognition system to transcribe participants' speech into text. We then gather opinions by inputting the transcribed text into ChatGPT as well as a prompt designed to solicit fluency evaluations. Subsequently, we extract embeddings from the speech, text, and opinions by the pretrained models. Finally, we use a classifier consisting of transformer blocks and linear layers to identify participants with this type of dementia. Experiments are conducted using the extensively used ADReSSo dataset. The results yield a maximum accuracy of 87.3% when speech, text, and opinions are used in conjunction. This finding suggests the potential of leveraging evaluation feedback from language models to address challenges in Alzheimer's disease recognition.