• Journal of the Institute of Convergence Signal Processing
• /
• v.4 no.4
• /
• pp.30-39
• /
• 2003

In this paper, we carried out the performance evaluation of HM-Net(Hidden Markov Network) speech recognition system for Korean speech databases. We adopted to construct acoustic models using the HM-Nets modified by HMMs(Hidden Markov Models), which are widely used as the statistical modeling methods. HM-Nets are carried out the state splitting for contextual and temporal domain by PDT-SSS(Phonetic Decision Tree-based Successive State Splitting) algorithm, which is modified the original SSS algorithm. Especially it adopted the phonetic decision tree to effectively express the context information not appear in training speech data on contextual domain state splitting. In case of temporal domain state splitting, to effectively represent information of each phoneme maintenance in the state splitting is carried out, and then the optimal model network of triphone types are constructed by in the parameter. Speech recognition was performed using the one-pass Viterbi beam search algorithm with phone-pair/word-pair grammar for phoneme/word recognition, respectively and using the multi-pass search algorithm with n-gram language models for sentence recognition. The tree-structured lexicon was used in order to decrease the number of nodes by sharing the same prefixes among words. In this paper, the performance evaluation of HM-Net speech recognition system is carried out for various recognition conditions. Through the experiments, we verified that it has very superior recognition performance compared with the previous introduced recognition system.

• Nuclear Engineering and Technology
• /
• v.50 no.8
• /
• pp.1217-1233
• /
• 2018

Following a surge of interest in multi-unit risk in the last few years, many recent studies have suggested methods for multi-unit probabilistic safety assessment (MUPSA) and addressed several related aspects. Most of the existing studies though focused on two-unit nuclear power plant (NPP) sites or used rather simplified probabilistic safety assessment (PSA) models to demonstrate the proposed approaches. When considering an NPP site with three or more units, some approaches are inapplicable or yield very conservative results. Since the number of such sites is increasing, there is a strong need to develop and validate practical approaches to the related MUPSA. This article provides several detailed approaches that are applicable to multi-unit Level 1 PSA for sites with up to six or more reactor units. To validate the approaches, a multi-unit Level 1 PSA model is developed and the site core damage frequency is estimated for each of four representative multi-unit initiators, as well as for the case of a simultaneous occurrence of independent single-unit initiators in multiple units. For this purpose, an NPP site with six identical OPR-1000 units is considered, with full-scale Level 1 PSA models for a specific OPR-1000 plant used as the base single-unit models.

• Nuclear Engineering and Technology
• /
• v.50 no.8
• /
• pp.1199-1209
• /
• 2018

After the Fukushima-Daiichi accident in 2011, the multi-unit risk, i.e., the risk due to several nuclear power plants (NPPs) in a site has become an important issue in several countries such as Korea, Canada, and China. However, the multi-unit risk has been discussed for a long time in the nuclear community before the Fukushima-Daiichi nuclear accident occurred. The regulatory authorities around the world and the international organizations had proposed requirements or guidelines to reduce the multi-unit risk. The concerns regarding the multi-unit risk can be summarized in the following three questions: How much the accident of an NPP in a site affects the safety of other NPPs in the same site? What is the total risk of a site with many NPPs? Will the risk of the simultaneous accidents at several NPPs in a site such as the Fukushima Daiichi accident be low enough? The multi-unit risk assessment (MURA) in an integrated framework is a practical approach to obtain the answers for the above questions. Even though there were few studies to assess the multi-unit risk before the Fukushima-Daiichi nuclear accident, there are still several issues to be resolved to perform the complete MURA. This article aims to provide an overview of the multi-unit risk issues and its assessment. We discuss the several critical issues in the current MURA to get useful insights regarding the multi-unit risk with the current state art of probabilistic safety assessment (PSA) technologies. Also, the qualitative answers for the above questions are addressed.

• Journal of Institute of Control, Robotics and Systems
• /
• v.10 no.10
• /
• pp.878-886
• /
• 2004

Uncertainties are inevitable in robotic assembly in unstructured environment since it is difficult to construct fixtures to guide motions of robots. This paper proposes an augmented Petri net and an algorithm to adapt the assembly model on-line during actual assembly process. The augmented Petri net identifies events using force and position information simultaneously. Unmodeled contact states are identified and incorporated into the model on-line. The proposed method increases the level of intelligence of the robot system by enhancing the autonomy. The proposed method is evaluated by simulation and experiments with L-type peg-in-hole assembly using a two-arm robot system.

• Nuclear Engineering and Technology
• /
• v.39 no.4
• /
• pp.249-256
• /
• 2007

Light water reactors operated under supercritical pressure conditions have been selected as one of the promising future reactor concepts to be studied by the Generation IV International Forum. Whereas the steam cycle of such reactors can be derived from modem fossil fired power plants, the reactor itself, and in particular the reactor core, still need to be developed. Different core design concepts shall be described here to outline the strategy. A first option for near future applications is a pressurized water reactor with $380^{\circ}C$ core exit temperature, having a closed primary loop and achieving 2% pts. higher net efficiency and 24% higher specific turbine power than latest pressurized water reactors. More efficiency and turbine power can be gained from core exit temperatures around $500^{\circ}C$, which require a multi step heat up process in the core with intermediate coolant mixing, achieving up to 44% net efficiency. The paper summarizes different core and assembly design approaches which have been studied recently for such High Performance Light Water Reactors.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2018.11a
• /
• pp.81-83
• /
• 2018

Fully Convolutional Network(FCN)은 기존의 방법보다 뛰어난 성능을 보였지만, FCN은 RGB 정보만을 사용하기 때문에 세밀한 예측이 필요한 장면에서는 다소 부족한 성능을 보였다. 이를 해결하기 위해 인코더-디코더 구조를 이용하여 RGB와 깊이의 멀티 모달을 활용하기 위한 FuseNet이 제안되었다. 하지만, FuseNet에서는 RGB와 깊이 브랜치 사이의 융합은 있지만, 인코더와 디코더 사이의 특징 지도를 융합하지 않는다. 본 논문에서는 FCN의 디코더 부분의 업샘플링 과정에서 이전 계층의 결과와 2배 업샘플링한 결과를 융합하는 스킵 레이어를 적용하여 FuseNet의 모달리티를 잘 활용하여 성능을 개선했다. 본 실험에서는 NYUDv2와 SUNRGBD 데이터 셋을 사용했으며, 전체 정확도는 각각 77%, 65%이고, 평균 IoU는 47.4%, 26.9%, 평균 정확도는 67.7%, 41%의 성능을 보였다.

• International Journal of High-Rise Buildings
• /
• v.10 no.1
• /
• pp.35-43
• /
• 2021

With the rapid urbanization and growing energy use intensity in the built environment, the glazed curtainwall has become ever more important in the architectural practice and environmental stewardship. Besides its energy efficiency roles, window has been an important transparent component for daylight penetration and a view-out for occupant satisfaction. In response to the climate crisis caused by the built environment, this research focuses on the study of net-zero energy retrofitting by using a new building integrated photovoltaic (BIPV) curtainwall as a sustainable alternative to conventional window systems. Design variables such as building orientations, climate zones, energy attributes of BIPV curtainwalls, and glazed area were studied, to minimize energy consumption and discomfort hours for three cities representing hot (Miami, FL), mixed (Charlotte, NC), and cold (Minneapolis, MN). Parametric analysis and Pareto solutions are presented to provide a comprehensive explanation of the correlation between design variables and performance objectives for net-zero energy retrofitting applications.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.621-623
• /
• 2023

본 연구는 당뇨병성 망막증의 자동 분류를 위해 딥러닝 모델을 활용한다. CLAHE 를 사용한 전처리로 이미지의 대비를 향상시켰으며, ResNet50 모델을 기반으로 한 전이학습을 통해 모델의 성능을 향상했다. 또한, 데이터의 불균형을 고려하여 정확도 뿐만 아니라 민감도와 특이도를 평가함으로써 모델의 분류 성능을 종합적으로 평가하였다. 실험 결과, 제안한 방법은 당뇨병성 망막증 분류 작업에서 높은 정확도를 달성하였으나, 양성 클래스의 식별에서 일부 한계가 있었다. 이에 데이터의 품질 개선과 불균형 데이터 처리에 초점을 맞춘 향후 연구 방향을 제시하였다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.592-595
• /
• 2024

엄청난 성능에도 불구하고, 심층 신경망은 예측결과에 대한 설명이 없는 블랙 박스로 작동한다는 비판을 받고 있다. 이러한 불투명한 표현은 신뢰성을 제한하고 모델의 대한 과학적 이해를 방해한다. 본 연구는 여러 개의 교사 신경망으로부터 설명 중심의 학생 신경망으로 지식 증류를 통해 해석 가능성을 향상시키는 것을 제안한다. 구체적으로, 인간이 정의한 개념 활성화 벡터 (CAV)를 통해 교사 모델의 개념 민감도를 방향성 도함수를 사용하여 계량화한다. 목표 개념에 대한 민감도 점수에 비례하여 교사 지식 융합을 가중치를 부여함으로써 증류된 학생 모델은 양호한 성능을 달성하면서 네트워크 논리를 해석으로 집중시킨다. 실험 결과, ResNet50, DenseNet201 및 EfficientNetV2-S 앙상블을 7 배 작은 아키텍처로 압축하여 정확도가 6% 향상되었다. 이 방법은 모델 용량, 예측 능력 및 해석 가능성 사이의 트레이드오프를 조화하고자 한다. 이는 모바일 플랫폼부터 안정성이 중요한 도메인에 걸쳐 믿을 수 있는 AI 의 미래를 여는 데 도움이 될 것이다.

• Proceedings of the Korea Multimedia Society Conference
• /
• 2004.05a
• /
• pp.387-390
• /
• 2004

통신망의 규모가 커지고 구조가 복잡해짐에 따라, 통신망의 성능을 최적화하여 사용자들이 요구하는 서비스 품질을 보장해 주는 성능관리의 기능이 절실히 요구되고 있다. 현재 성능관리의 주요 기능이 되는 트래픽에 대한 분석을 위해서 넷플로우(NetFlow), RMON, 그리고 패킷을 캡처하는 방법이 쓰이고 있지만 통합적인 관점의 해결책은 되지 못한다. 본 논문에서는 다양한 전송기술(Multi-technology), 다양한 장치 제조사(Multi-vender) 장비들의 성능관리를 가능케 할 수 있도록 통합적인 통신망 성능관리 구조를 제시하고, 그에 따라 성능관리 시스템을 설계하고 구현하였다.