• 한국신뢰성학회지:신뢰성응용연구
• /
• 제13권4호
• /
• pp.229-239
• /
• 2013

This paper reports a manner to use a Bayesian approach to derive the optimal replacement policy. In order to produce a system with minimal repair warranty, a replacement model with the extended warranty is considered. Within the warranty period, the failed system is minimally repaired by the manufacturer at no cost to the end-user. The failure time is assumed to follow a Weibull distribution with unknown parameters. The expected cost rate per unit time, from the end-user's viewpoints, is induced by the Bayesian approach, and the optimal replacement policy to minimize the cost rate is proposed. Finally, a numerical example illustrating to derive the optimal replacement policy based on the Bayesian approach is described.

• 한국음향학회지
• /
• 제38권5호
• /
• pp.614-620
• /
• 2019

본 논문은 전역 스타일 토큰(Global Style Token, GST)을 사용하는 종단 간(end-to-end) 감정 음성 합성 시스템의 성능을 높이기 위해 각 감정의 스타일 벡터를 효과적으로 결정하는 방법을 제안한다. 기존 방법은 각 감정을 표현하기 위해 한 개의 대푯값만을 사용하므로 감정 표현의 풍부함 측면에서 크게 제한된다. 이를 해결하기 위해 본 논문에서는 k-평균 알고리즘을 사용하여 다수의 대표 스타일을 추출하는 방법을 제안한다. 청취 평가를 통해 제안 방법을 이용해 추출한 각 감정의 대표 스타일이 기존 방법에 비해 감정 표현 정도가 뛰어나며, 감정 간의 차이를 명확히 구별할 수 있음을 보였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권6호
• /
• pp.1692-1705
• /
• 2024

In this paper, we present a method that integrates a Grammar Transducer as an external language model to enhance the accuracy of the pre-trained Korean End-to-end (E2E) Automatic Speech Recognition (ASR) model. The E2E ASR model utilizes the Connectionist Temporal Classification (CTC) loss function to derive hypothesis sentences from input audio. However, this method reveals a limitation inherent in the CTC approach, as it fails to capture language information from transcript data directly. To overcome this limitation, we propose a fusion approach that combines a clause-level n-gram language model, transformed into a Weighted Finite-State Transducer (WFST), with the E2E ASR model. This approach enhances the model's accuracy and allows for domain adaptation using just additional text data, avoiding the need for further intensive training of the extensive pre-trained ASR model. This is particularly advantageous for Korean, characterized as a low-resource language, which confronts a significant challenge due to limited resources of speech data and available ASR models. Initially, we validate the efficacy of training the n-gram model at the clause-level by contrasting its inference accuracy with that of the E2E ASR model when merged with language models trained on smaller lexical units. We then demonstrate that our approach achieves enhanced domain adaptation accuracy compared to Shallow Fusion, a previously devised method for merging an external language model with an E2E ASR model without necessitating additional training.

• 대한기관식도과학회지
• /
• 제3권1호
• /
• pp.61-69
• /
• 1997

A total of 55 patients underwent surgical managements for postintubation tracheal stenosis from July 1975 through March 1997. All but 8 had received ventilatory assistance. The patients had S cuff lesions, 17 stoma lesions, 7 at both levels, 5 at subglottic lesions. Thirty two patients underwent the sleeve tracheal resection and end-to-end anastomosis. Five patients performed a wedge resection and end-to-end anastomosis. Twenty two patients received the Montgomery T-tube for relief of airway obstruction. Simple excision of granulation tissue was done in 7 patients. Rethi procedures(anterior division of cricoid cartilage, partial wedge resection of lower thyroid cartilage and T-tube molding) were performed in 2 subglottic stenosis patients. And the other subglottic patient was received permanent tracheal fenestration at 1975. The tracheoesophageal fistula patient was done sleeve tracheal resection and end-to-end anastomosis with interrupted double layer closure of esophageal fistula site. Cervical approach was used in 49 cases, cervicomediastinal in 13 cases and median stemotomy In 6 cases. Techniques for obtaining tension-free anastomosis included a cervical neck flexion(15-30$^{\circ}$) in all sleeve resection patients and laryngeal release in one. The length of resection was 1.5 to 5.0 on A total of 41 patients(74.5%) had good(24 patients) or satisfactory(17 patients) results. But in ten cases, the restenosis of anastomosis site which is the most common complication was developed Two of them underwent a second reconstruction and 8 patients required T-tube insertion for airway maintenance. Three patients(5.4%) died. The causes of death were tracheo-innominate artery fistula(2) and sudden obstruction of airway(1).

• Structural Engineering and Mechanics
• /
• 제31권3호
• /
• pp.297-314
• /
• 2009

When the temperature of a structure varies, there is a tendency to produce changes in the shape of the structure. The resulting actions may be of considerable importance in the analysis of the structures having non-prismatic members. Therefore, this study aimed to investigate the modeling, analysis and behavior of the non-prismatic members subjected to temperature changes with the aid of finite element modeling. The fixed-end moments and fixed-end forces of such members due to temperature changes were computed through a comprehensive parametric study. It was demonstrated that the conventional methods using frame elements can lead to significant errors, and the deviations can reach to unacceptable levels for these types of structures. The design formulas and the dimensionless design coefficients were proposed based on a comprehensive parametric study using two-dimensional plane-stress finite element models. The fixed-end actions of the non-prismatic members having parabolic and straight haunches due to temperature changes can be determined using the proposed approach without necessitating a detailed finite element model solution. Additionally, the robust results of the finite element analyses allowed examining the sources and magnitudes of the errors in the conventional analysis.

• 디지털콘텐츠학회 논문지
• /
• 제18권1호
• /
• pp.93-99
• /
• 2017

Long Short Term Memory (LSTM) Recurrent Neural Network (RNN)를 이용한 hybrid 방법은 음성 인식률을 크게 향상시켰다. Hybrid 방법에 기반한 음향모델을 학습하기 위해서는 Gaussian Mixture Model (GMM)-Hidden Markov Model (HMM)로부터 forced align된 HMM state sequence가 필요하다. 그러나, GMM-HMM을 학습하기 위해서 많은 연산 시간이 요구되고 있다. 본 논문에서는 학습 속도를 향상하기 위해, LSTM RNN 기반 한국어 음성인식을 위한 end-to-end 방법을 제안한다. 이를 구현하기 위해, Connectionist Temporal Classification (CTC) 알고리즘을 제안한다. 제안하는 방법은 기존의 방법과 비슷한 인식률을 보였지만, 학습 속도는 1.27 배 더 빨라진 성능을 보였다.

• Structural Engineering and Mechanics
• /
• 제56권6호
• /
• pp.899-916
• /
• 2015

Upgrading or strengthening of existing reinforced concrete (RC) infrastructure is an emerging demand nowadays. Near Surface Mounted (NSM) technique is very promising approach for flexural strengthening of RC members. However, premature failure such as concrete cover separation failure have been a main concern in utilizing this technique. In this study, U-wrap end anchorage with carbon fiber reinforced polymer (CFRP) fabrics is proposed to eliminate the concrete cover separation failure. Experimental programs were conducted to the consequence of U-wrap end anchorage on the flexurally strengthened RC beams with NSM-steel. A total of eight RC rectangular beam specimens were tested. One specimen was kept unstrengthened as a reference; three specimens were strengthened with NSM-steel bars and the remaining four specimens were strengthened with NSM-steel bars and U-wrap end anchorage using CFRP fabrics. A 3D non-linear finite element model (FEM) was developed to simulate the flexural response of the tested specimens. It is revealed that NSM-steel (with and without end-anchors) significantly improved the flexural strength; moreover decreased deflection and strains compared with reference specimen. Furthermore, NSM-steel with end anchorage strengthened specimens revealed the greater flexural strength and improve failure modes (premature to flexure) compared with the NSM-steel without end anchorage specimens. The results also ensured that the U-wrap end anchorage completely eliminate the concrete cover separation failure.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1992년도 한국자동제어학술회의논문집(국제학술편); KOEX, Seoul; 19-21 Oct. 1992
• /
• pp.592-598
• /
• 1992

For a high bandwidth, accurate end of arm motion control with good disturbance rejection, the, Momentum Management Approach to Motion control (MMAM) is proposed. The MMAM is a kind of position control technique that uses inertial forces, applied at or near the end of arm to achieve, high bandwidth and accuracy in movement and in the face of force disturbances. To prove the concept of MMAM, the, end point, control of a flexible manipulator is considered. For this purpose, a flexible beam is mounted on the x-y table, and the MMAM actuator is attached on the top of the flexible beam. A mathematical model is developed for the flexible, beam being controlled by the, MMAM actuator and slide base DC motor. A system identification method is applied to estimate some system parameters in the, model which can not be determined because of the complexity of the mechanism. For the end point, control of the. flexible beam, the, optimal linear output feedback control is introduced.

• 인터넷정보학회지
• /
• 제14권3호
• /
• pp.58-67
• /
• 2013

An essential prerequisite for negotiating governance of 'internet,' a world of 'end to end' nature, is to guarantee end-users' participation. Internet Governance Forum (IGF), the outcome of World Summit on the Information Society(WSIS) Tunis Agenda, has played an important role as a representative mulitistakeholder - based governance model, which puts emphasis on wide range participation of stakeholders. Notwithstanding multistakeholderism materialized in IGF has been challenged, IGF itself is a unique place for participation to broad internet governance discourse, as all voices get hearing grounded on openness, inclusion, and transparency, quite distinct from the multilateral negotiation based model such as 'Conference on Cyberspace.' Therefore civil society which represents end-users' interest, as a crucial stakeholder of internet governance, should establish more strategic and coordinated approach to IGF itself as well as reformation discourse of internet governance. In this regard, civil society groups of Korea, began with hosting workshops and Open Forum at the 2013 IGF in Bali, Indonesia. This report also describes the detailed activities of Korean civil society group in 2013 IGF.

• 한국생산제조학회지
• /
• 제19권4호
• /
• pp.498-503
• /
• 2010

The main purpose of this study strongly concerned micro machining error estimation by using FEM analysis of tool deflection shapes in micro flat end-milling process. For the precision micro flat end-milling process, analysis of micro cutting errors is mandatory. In general, tool deflection is a major factor which causes cutting error and limits realization of the high-precision cutting process. Especially, in micro end-milling process, micro tool deflection generates very serious problems in contrast to macro tool deflection. Methods which deal with compensation of cutting error by tool deflection in macro end-milling process have been studied plentifully but, few researches transact with micro scaled cutting tool deflection in micro cutting process. Therefore, the trend of micro tool deflection was estimated by using FEM analysis in this paper. Cutting forces were acquired by micro dynamometer and these were utilized in FEM analysis. In order to verify FEM analysis results, micro machining processes were carried out and real machined profiles were compared with FEM results. Finally through the proposed approach well suited FEM results were obtained.