• 마이크로전자및패키징학회지
• /
• 제30권4호
• /
• pp.69-78
• /
• 2023

산업 인공지능의 발달과 함께 반도체의 수요가 크게 증가하고 있다. 시장 수요에 대응하기 위해 패키징 공정에서 자동 결함 검출의 중요성 역시 증가하고 있다. 이에 따라, 패키지의 자동 불량 검사를 위한 딥러닝 기반의 방법론들의 연구가 활발히 이루어 지고 있다. 딥러닝 기반의 모델은 학습을 위해서 대량의 고해상도 데이터를 필요로 하나, 보안이 중요한 반도체 분야의 특성상 관련 데이터의 공유 및 레이블링이 쉽지 않아 모델의 학습이 어려운 한계를 지니고 있다. 또한 고해상도 이미지를 생성하기 위해 상당한 컴퓨팅 자원이 요구되는데, 본 연구에서는 분할정복 접근법을 통해 적은 컴퓨팅 자원으로 딥러닝 모델 학습을 위한 충분한 양의 데이터를 확보하는 방법을 소개한다. 제안된 방법은 높은 해상도의 이미지를 분할하고 각 영역에 조건 레이블을 부여한 후, 독립적인 부분 영역과 경계를 학습시켜, 경계 손실이 일관적인 이미지를 생성하도록 유도한다. 이후, 분할된 이미지를 하나로 통합하여, 최종적으로 모델이 고해상도의 이미지를 생성하도록 구성하였다. 실험 결과, 본 연구를 통해 증강된 이미지들은 높은 효율성, 일관성, 품질 및 범용성을 보였다.

• 한국인터넷방송통신학회논문지
• /
• 제23권6호
• /
• pp.15-20
• /
• 2023

본 논문은 캐릭터 웹드라마에 적합한 간접광고 제품을 추천하는 인공지능(AI) 시스템 알고리즘 개발에 관한 연구이다. 본 연구는 웹드라마의 대사 작성에 있어 그에 어울리는 간접광고 제품을 추천함으로써 시청자의 콘텐츠 몰입도를 높이고, 드라마의 스토리를 보다 깊게 이해하는 데 도움을 주는 것을 목표로 한다. 본 연구에서는 자연어처리 모델 인 GPT를 활용하여 대사, 줄거리를 분석하고, 분석 결과를 바탕으로 소품형, 배경형 등 두 가지 유형의 간접광고 제품 추천 시스템을 개발한다. 이를 통해 웹드라마의 스토리에 부합하는 제품을 적절히 배치함으로써 간접광고가 자연스럽게 노출될 수 있도록 하고, 그로 인해 시청자들의 몰입도가 증가하며, 상품 홍보의 효과 또한 높인다. 숨겨진 뜻이나 문화적 뉘앙스를 완벽하게 이해하기 어려운 인공지능 모델의 한계와 학습에 필요한 충분한 데이터 확보가 어렵다는 한계가 있다. 그러나 본 연구는 AI가 창작물 제작에 어떻게 기여할 수 있는지에 대한 새로운 인사이트를 제공하고, 창의적 산업 분야에서 자연어 처리 모델의 활용 가능성을 넓히는 중요한 발판이 될 것이다.

• 보건행정학회지
• /
• 제33권4호
• /
• pp.479-488
• /
• 2023

Background: By applying the suggested criteria for needs-based chronic medical care and long-term care delivery system for the elderly, the current status of delivery system was identified and regional delivery systems were categorized according to quantity and quality of delivery system. Methods: National claims data were used for this study. All claims data of medical and long-term care uses by the elderly and all claims data from long-term care hospitals and nursing homes in 2016 were analyzed to categorize the regional medical and long-term care delivery system. The current status of the delivery system with a high possibility of transition to a needs-based appropriate delivery system was identified. The necessary and actual amount of regional supply was calculated based on their needs, and the structure of delivery systems was evaluated in terms of the needs-based quality of the system. Finally, all regions were categorized into 15 types of medical and care delivery systems for the elderly. Results: Of the total 55 regions, 89.1% of regions had an oversupply of elderly medical and care services compared to the necessary supply based on their needs. However, 69.1% of regions met the criteria for less than two types of needs groups, and 21.8% of regions were identified as regions where the numbers of institutions or regions with a high possibility of transition to an appropriate delivery system were below the average levels for all four needs groups. Conclusion: In order to establish an appropriate community-based integrated elderly care system, it is necessary to analyze the characteristics of the regional delivery system categories and to plan a needs-based delivery system regionally.

• 로봇학회논문지
• /
• 제18권4호
• /
• pp.385-391
• /
• 2023

In the elderly population, sarcopenia occurs due to physical aging, leading to movement restrictions and loss of function. This results in dependence on daily activities and limitations in participation, ultimately decreasing the overall quality of life. In this study, we propose an algorithm designed to enable patients with sarcopenia to perform sit-to-stand and stand-to-sit movements seamlessly in their daily lives. The algorithm incorporates a wearable robot for muscle support and includes algorithms for standing and seated muscle strength support. To validate the algorithm's performance, EMG sensors were attached to the Rectus Femoris and Biceps Femoris muscles. The participants underwent two scenarios: one without wearing the device and one with the device providing muscle strength support, performing sit-to-stand and stand-to-sit motions for one minute in each case. The results showed a 16% increase in the EMG peak value of the Rectus Femoris muscle during standing motion (p=0.009). On the right side, there was a roughly 20% decrease (p=0.018) during standing and a 21% decrease (p=0.014) during sitting motion. In the future, we aim to gather additional data to further refine the algorithm. Our goal is to develop an optimal muscle strength support algorithm based on this data, making it applicable for real-life use by patients with sarcopenia.

• 문화기술의 융합
• /
• 제9권6호
• /
• pp.703-708
• /
• 2023

본 연구에서는 인공지능 기술을 이용하는 사용자를 이해하는 계획된 행동이론에 집중하여 다양한 방면에 적용되고 있는 가상개인비서 이용을 살펴보았다. 먼저, 사용자의 신념과 태도, 지각된 가치 및 위험성이 가상개인비서 지속사용의도에 미치는 영향을 알아본 결과, (1) 가상개인비서 이용에 대한 태도와 주관적 규범, 지각된 행동 통제, 그리고 지각된 가치는 지속사용의도에 정적으로 유의미한 영향을 미쳤다. 다음으로 (2) 사용자의 신념 및 태도와 지각된 가치 및 위험성이 가상개인비서에 대한 구전의도에 미치는 영향에 대하여 알아본 결과, 지각된 위험성을 제외한 모든 변수들(사용자 태도와 주관적 규범, 지각된 행동 통제, 그리고 지각된 가치)이 구전의도에 정적으로 유의미한 영향을 미쳤다. 본 연구의 결과는 사용자의 신념 및 태도에서 나아가 사용자가 지각하는 가치와 위험성이 가상개인비서 이용에 미치는 영향을 분석하여, 폭발적으로 성장하고 있는 인공지능 시장에 다양한 실무적 및 이론적 함의를 제공할 것이라 기대한다.

• Child Health Nursing Research
• /
• 제29권4호
• /
• pp.290-299
• /
• 2023

Purpose: Artificial intelligence (AI) has had a profound impact on humanity; in particular, chatbots have been designed for interactivity and applied to many aspects of daily life. Chatbots are also regarded as an innovative modality in nursing education. This study aimed to identify nursing students' awareness of using chatbots and factors influencing their usage intention. Methods: This study, which employed a descriptive design using a self-reported questionnaire, was conducted at three university nursing schools located in Seoul, South Korea. The participants were 289 junior and senior nursing students. Data were collected using self-reported questionnaires, both online via a Naver Form and offline. Results: The total mean score of awareness of using chatbots was 3.49±0.61 points out of 5. The mean scores of the four dimensions of awareness of using chatbots were 3.37±0.60 for perceived value, 3.66±0.73 for perceived usefulness, 3.83±0.73 for perceived ease of use, and 3.36±0.87 for intention to use. Significant differences were observed in awareness of using chatbots according to satisfaction with nursing (p<.001), effectiveness of using various methods for nursing education (p<.001), and interest in chatbots (p<.001). The correlations among the four dimensions ranged from .52 to .80. In a hierarchical regression analysis, perceived value (β=.45) accounted for 60.2% of variance in intention to use. Conclusion: The results suggest that chatbots have the potential to be used in nursing education. Further research is needed to clarify the effectiveness of using chatbots in nursing education.

• 한국IT서비스학회지
• /
• 제22권5호
• /
• pp.99-108
• /
• 2023

As the computerization of hospitals becomes more advanced, security issues regarding data generated from various medical devices within hospitals are gradually increasing. For example, because hospital data contains a variety of personal information, attempts to attack it have been continuously made. In order to safely protect data from external attacks, each hospital has formed an internal team to continuously monitor whether the computer network is safely protected. However, there are limits to how humans can monitor attacks that occur on networks within hospitals in real time. Recently, artificial intelligence models have shown excellent performance in detecting outliers. In this paper, an experiment was conducted to verify how well an artificial intelligence model classifies normal and abnormal data in network traffic data generated from medical devices. There are several models used for outlier detection, but among them, Random Forest and Tabnet were used. Tabnet is a deep learning algorithm related to receive and classify structured data. Two algorithms were trained using open traffic network data, and the classification accuracy of the model was measured using test data. As a result, the random forest algorithm showed a classification accuracy of 93%, and Tapnet showed a classification accuracy of 99%. Therefore, it is expected that most outliers that may occur in a hospital network can be detected using an excellent algorithm such as Tabnet.

• Imaging Science in Dentistry
• /
• 제53권4호
• /
• pp.271-281
• /
• 2023

Purpose: The objective of this scoping review was to investigate the applicability and performance of various convolutional neural network (CNN) models in tooth numbering on panoramic radiographs, achieved through classification, detection, and segmentation tasks. Materials and Methods: An online search was performed of the PubMed, Science Direct, and Scopus databases. Based on the selection process, 12 studies were included in this review. Results: Eleven studies utilized a CNN model for detection tasks, 5 for classification tasks, and 3 for segmentation tasks in the context of tooth numbering on panoramic radiographs. Most of these studies revealed high performance of various CNN models in automating tooth numbering. However, several studies also highlighted limitations of CNNs, such as the presence of false positives and false negatives in identifying decayed teeth, teeth with crown prosthetics, teeth adjacent to edentulous areas, dental implants, root remnants, wisdom teeth, and root canal-treated teeth. These limitations can be overcome by ensuring both the quality and quantity of datasets, as well as optimizing the CNN architecture. Conclusion: CNNs have demonstrated high performance in automated tooth numbering on panoramic radiographs. Future development of CNN-based models for this purpose should also consider different stages of dentition, such as the primary and mixed dentition stages, as well as the presence of various tooth conditions. Ultimately, an optimized CNN architecture can serve as the foundation for an automated tooth numbering system and for further artificial intelligence research on panoramic radiographs for a variety of purposes.

• Structural Engineering and Mechanics
• /
• 제90권1호
• /
• pp.19-26
• /
• 2024

In recent years, an increasing number of experimental studies have shown that the practical application of mature active control systems requires consideration of robustness criteria in the design process, including the reduction of tracking errors, operational resistance to external disturbances, and measurement noise, as well as robustness and stability. Good uncertainty prediction is thus proposed to solve problems caused by poor parameter selection and to remove the effects of dynamic coupling between degrees of freedom (DOF) in nonlinear systems. To overcome the stability problem, this study develops an advanced adaptive predictive fuzzy controller, which not only solves the programming problem of determining system stability but also uses the law of linear matrix inequality (LMI) to modify the fuzzy problem. The following parameters are used to manipulate the fuzzy controller of the robotic system to improve its control performance. The simulations for system uncertainty in the controller design emphasized the use of acceleration feedback for practical reasons. The simulation results also show that the proposed H∞ controller has excellent performance and reliability, and the effectiveness of the LMI-based method is also recognized. Therefore, this dynamic control method is suitable for seismic protection of civil buildings. The objectives of this document are access to adequate, safe, and affordable housing and basic services, promotion of inclusive and sustainable urbanization, implementation of sustainable disaster-resilient construction, sustainable planning, and sustainable management of human settlements. Simulation results of linear and non-linear structures demonstrate the ability of this method to identify structures and their changes due to damage. Therefore, with the continuous development of artificial intelligence and fuzzy theory, it seems that this goal will be achieved in the near future.

• 한국인터넷방송통신학회논문지
• /
• 제24권1호
• /
• pp.141-147
• /
• 2024

최근 개인의 성향을 반영한 맞춤형 서비스가 각광 받고 있다. 이와 관련하여 개인의 개성을 인식하고 활용하고자 하는 연구가 지속적으로 이루어지고 있다. 각 개인의 개성을 인식하고 평가하는 방법은 다수가 있지만, OCEAN 모델이 대표적으로 사용된다. OCEAN 모델로 각 개인의 개성을 인식할 때 언어적, 준언어적, 비언어적 정보를 이용하는 멀티 모달리티 기반 인공지능 모델이 사용될 수 있다. 본 논문에서는 비언어적 정보인 사용자의 표정을 기반으로 OCEAN을 인식하는 인공지능 모델에서 영상 데이터에서 얼굴 영역을 추출할 때 지정하는 얼굴 영역 여유값(Margin)에 따른 개성 인식 모델 정확도 성능을 분석한다. 실험에서는 2D Patch Partition, R2plus1D, 3D Patch Partition, 그리고 Video Swin Transformer에 기반한 개성 인식 모델을 사용하였다. 얼굴 영역 추출 시 여유값을 60으로 사용했을 때 1-MAE 성능이 0.9118로 가장 우수하였다. 따라서 개성 인식 모델의 성능을 최적화하기 위해서는 적절한 여유값을 설정해야 함을 확인하였다.