• 한국음향학회지
• /
• 제39권5호
• /
• pp.447-453
• /
• 2020

본 논문은 딥러닝 기법을 활용하여 음성신호로부터 효율적인 화자 벡터를 추출하는 시스템을 제안한다. 음성신호에는 발화내용, 감정, 배경잡음 등과 같이 화자의 특징과는 관련이 없는 정보들이 포함되어 있다는 점에 착안하여 제안 방법에서는 추출된 화자 벡터에 화자의 특징과 관련된 정보는 가능한 많이 포함되고, 그렇지 않은 비화자 정보는 최소화될 수 있도록 학습을 진행한다. 특히, 오토-인코더 구조의 부호화 기가 두 개의 임베딩 벡터를 추정하도록 하고, 효과적인 손실 함수 조건을 두어 각 임베딩이 화자 및 비화자 특징만 각각 포함할 수 있도록 하는 효과적인 화자 정보 분리(disentanglement)방법을 제안한다. 또한, 화자 정보를 유지하는데 도움이 되는 생성적 적대 신경망(Generative Adversarial Network, GAN)에서 활용되는 판별기 구조를 도입함으로써, 디코더의 성능을 향상시킴으로써 화자 인식 성능을 보다 향상시킨다. 제안된 방법에 대한 적절성과 효율성은 벤치마크 데이터로 사용되고 있는 Voxceleb1에 대한 동일오류율(Equal Error Rate, EER) 개선 실험을 통하여 규명하였다.

• 로봇학회논문지
• /
• 제15권2호
• /
• pp.197-204
• /
• 2020

In this paper, we present a learning platform for robotic grasping in real world, in which actor-critic deep reinforcement learning is employed to directly learn the grasping skill from raw image pixels and rarely observed rewards. This is a challenging task because existing algorithms based on deep reinforcement learning require an extensive number of training data or massive computational cost so that they cannot be affordable in real world settings. To address this problems, the proposed learning platform basically consists of two training phases; a learning phase in simulator and subsequent learning in real world. Here, main processing blocks in the platform are extraction of latent vector based on state representation learning and disentanglement of a raw image, generation of adapted synthetic image using generative adversarial networks, and object detection and arm segmentation for the disentanglement. We demonstrate the effectiveness of this approach in a real environment.

• Nuclear Engineering and Technology
• /
• 제55권8호
• /
• pp.2854-2863
• /
• 2023

Objective: To present a hybrid approach that incorporates a constrained beam-hardening estimator (CBHE) and deep learning (DL)-based post-refinement for metal artifact reduction in dental cone-beam computed tomography (CBCT). Methods: Constrained beam-hardening estimator (CBHE) is derived from a polychromatic X-ray attenuation model with respect to X-ray transmission length, which calculates associated parameters numerically. Deep-learning-based post-refinement with an artifact disentanglement network (ADN) is performed to mitigate the remaining dark shading regions around a metal. Artifact disentanglement network (ADN) supports an unsupervised learning approach, in which no paired CBCT images are required. The network consists of an encoder that separates artifacts and content and a decoder for the content. Additionally, ADN with data normalization replaces metal regions with values from bone or soft tissue regions. Finally, the metal regions obtained from the CBHE are blended into reconstructed images. The proposed approach is systematically assessed using a dental phantom with two types of metal objects for qualitative and quantitative comparisons. Results: The proposed hybrid scheme provides improved image quality in areas surrounding the metal while preserving native structures. Conclusion: This study may significantly improve the detection of areas of interest in many dentomaxillofacial applications.

• Macromolecular Research
• /
• 제12권1호
• /
• pp.141-143
• /
• 2004

We have investigated the dynamic mechanical behavior of ultra-high molecular weight polyethylene (UHMWPE) irradiated with varying doses of gamma rays. A relaxation peak in the loss factor curve, which has not been reported previously in the literature, is observed at a temperature above the crystal melting temperature. The peak is unique to UHMWPE and appears to be related to the high degree of entanglement. Because the temperature and intensity of the peak are reduced by irradiation-induced chain scission and crosslinking, respectively, we believe that the peak is associated with disentanglement relaxation. The behavior of the storage modulus in the melt state agrees with the classical theory of rubber elasticity.

• 고전문학과교육
• /
• 제15호
• /
• pp.303-327
• /
• 2008

The purpose of this paper is the analysis of to point its structure and meaning deduced from the rite of passage. In the light of the rite of passage, could be retrieved its value as one of significant works among the works of Gangho-sijo. is composed of nineteen pieces of monotype sijo. The first piece is a kind of prologue. The other pieces are divided into three parts which correspond with the three stages of the rite of passage-separation, transition, incorporation. As a result of this analysis, we can find out the writer's cognitive change according to the schema of the rite of passage. It becomes clear that the space of Gangho means the sacred space for mental culture rather than a separate space for escape. And considering the withdrawal for Gangho as the transition process of the rite of passage, gives a description of the disentanglement of conflict and a incorporation into the superior position of spirit.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권6호
• /
• pp.2115-2127
• /
• 2021

Although significant progress has been made in synthesizing visually realistic face images by Generative Adversarial Networks (GANs), there still lacks effective approaches to provide fine-grained control over the generation process for semantic facial attribute editing. In this work, we propose a novel cross channel self-attention based generative adversarial network (CCA-GAN), which weights the importance of multiple channels of features and archives pixel-level feature alignment and conversion, to reduce the impact on irrelevant attributes while editing the target attributes. Evaluation results show that CCA-GAN outperforms state-of-the-art models on the CelebA dataset, reducing Fréchet Inception Distance (FID) and Kernel Inception Distance (KID) by 15~28% and 25~100%, respectively. Furthermore, visualization of generated samples confirms the effect of disentanglement of the proposed model.

• 부모자녀건강학회지
• /
• 제5권1호
• /
• pp.62-74
• /
• 2002

The purpose of this study was to explore bringing-up process of mothers with first baby in the context of Korean culture. The field survey was performed with four primiparae during four months from September, 2000 and grounded theory approach was used for data analysis. The results indicated that there were 27 primary categories and 12 secondary categories. Unclearness was extracted as an expression of the core category of bringing-up process. The subjects was faced with care of their baby after delivery and experienced formation and deepening of unclearness. After that, they also went through expression and disentanglement of complication caused by unclearness. It is found that the core concept of this study, the unclereness("Magmagham" in Korean) means a confusion, a burden, a helplessness and a hardness. New mothers make an effort to resolve this situation with strong motherhood in their mind. By supports from significant others, they are empowered and some mothers get depressive feelings. The findings of this study are contributed for nurses to understand new mothers in developmental crisis. We suggested that the nursing intervention strategies including learning baby care skills and preparing emotionally a motherhood must be given new mother during pregnancy period.

• 한국과학교육학회지
• /
• 제30권8호
• /
• pp.933-952
• /
• 2010

이 연구는 과학 관련 사회적 쟁점의 온라인 토론의 특징을 드러내고 이를 포스트모더니즘의 관점에서 해석하여 그 시사점을 제공하고자 한다. 연구 내용은 광우병과 관련된 과학적 주장 및 사회, 경제, 정치적 쟁점 간의 연결 관계를 보여주며 텍스트의 독창성, 저자와 독자의 관계, 발화 목적에 따른 특징을 드러내는 것이다. 이에 따라 본 연구는 온라인을 통한 사회적 논쟁의 대표적 사례인, 2008년 수입 쇠고기 논쟁을 중심으로 포털 사이트의 토론 게시판 66개의 글을 분석하였다. 연구 결과, 텍스트는 광우병과 관련된 5가지 과학적 논쟁 - 광우병의 발병원인, 변형 프리온의 검출 위치, 변형 프리온 섭취량과 인간 광우병의 발병 관계, 광우병에 대한 한국인의 유전적 취약 여부, 치매환자 증가와 광우병 사이의 상관관계 - 을 담고 있었다. 각 주장은 국내 소고기 시장의 파급 효과, 수입소 위생 및 사육환경, 무역 보복 가능성 등 다양한 사회적, 경제적, 정치적 문제가 얽혀 있었다. 독창성의 관점에서 텍스트는 짜깁기와 '펌', 독자에 의한 간섭을 통해 저자의 권한이 약화되고 집단적 특징의 글쓰기를 보여주었다. 나아가 저자와 독자 간의 거리를 좁히기 위해 저자의 정체성이나 개인적인 이야기들이 텍스트 속에 소개되었다. 발화 목적의 관점에서는 체계적인 논리와 정보의 전달보다는 감정 표현이나 행동의 변화가 중심이 되는 경향이 나타났다.

• 한국정밀공학회지
• /
• 제3권1호
• /
• pp.40-49
• /
• 1986

Crack, craze and void are common defects which may be found in the bulk of polymeric materials such as either themoplastics or thermosets. The healing phenomena, autohesion, of these defects are known to be a intrinsic material property of various polymeric materials. However, only a few experimental and theoretical investigations on crack, void and craze healing phenomena for various polymeric materials have been reported up to date [1, 2, 3]. This may be partly due to the complications of healing processes and lacking of appropriate theoretical developments. Recently, some investigators have been urged to study the healing phenomena of various polymenic materials since the significance of the use of polymer based alloys or composites has been raised in terms of specific strength and energy saving. In the earlier published reports [1, 2, 3, 4], the crack and void healing velocity, healing toughness and some other healing mechanical and physical properties were measured experimentally and compared with predicted values by utilizing a simple model such as the reptation model under some resonable assumptions. It seems, however, that the general acceptance of the proposed modeling analyses is yet open question. The crack healing processes seem to be complicate and highly dependent on the state of virgin material in terms of mechanical and physical properties. Furthermore, it is also strongly dependent on the histories of crack, craze and void development including fracture suface morphology, the shape of void and the degree of disentanglement of fibril in the craze. The rate of crack healing may be a function of environmental factors such as healing temperature, time and pressure which gives different contact configurations between two separated surfaces. It seems to be reasonable to assume that the crack healing processes may be divided in several distinguished steps like stress relaxation with molecular chain arrangement, surface contact (wetting), inter- diffusion process and com;oete healing (to obtain the original strength). In this context, it is likely that we no longer have to accept the limitation of cumulative damage theories and fatigue life if it is probable to remove the defects such as crack, craze and void and to restore the original strength of polymers or polymer based compowites by suitable choice of healing histories and methods. In this paper, we wish to present a very simple and intuitive theoretical model for the prediction of healed fracture toughness of cracked or defective polymeric components. The central idea of this investigation, thus, may be the modeling of behavior of chain molecules under healing conditions including the effects of chain scission on the healing processes. The validity of this proposed model will be studied by making comparisons between theoretically predicted values and experimentally determined results in near future and will be reported elsewhere.