• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.1
• /
• pp.27-32
• /
• 2019

Research that implements the classification process through Deep Learning algorithm, one of the outstanding human abilities, includes a unimodal model, a multi-modal model, and a multi-modal method using music videos. In this study, the results were better by suggesting a system to analyze each song's spectrum into short samples and vote for the results. Among Deep Learning algorithms, CNN showed superior performance in the category of music genre compared to RNN, and improved performance when CNN and RNN were applied together. The system of voting for each CNN result by Deep Learning a short sample of music showed better results than the previous model and the model with Softmax layer added to the model performed best. The need for the explosive growth of digital media and the automatic classification of music genres in numerous streaming services is increasing. Future research will need to reduce the proportion of undifferentiated songs and develop algorithms for the last category classification of undivided songs.

• The Journal of Society for e-Business Studies
• /
• v.27 no.2
• /
• pp.37-64
• /
• 2022

As the social acceptance of artificial intelligence increases, the number of cases of applying machine learning methods to companies is also increasing. Technical factors such as accuracy and interpretability have been the main criteria for selecting machine learning methods. However, the success of implementing machine learning also affects management factors such as IT departments, operation departments, leadership, and organizational culture. Unfortunately, there are few integrated studies that understand the success factors of machine learning selection in which technical and management factors are considered together. Therefore, the purpose of this paper is to propose and empirically analyze a technology-management integrated model that combines task-tech fit, IS Success Model theory, and John Rice's algorithm selection process model to understand machine learning selection within the company. As a result of a survey of 240 companies that implemented machine learning, it was found that the higher the algorithm quality and data quality, the higher the algorithm-problem fit was perceived. It was also verified that algorithm-problem fit had a significant impact on the organization's innovation and productivity. In addition, it was confirmed that outsourcing and management support had a positive impact on the quality of the machine learning system and organizational cultural factors such as data-driven management and motivation. Data-driven management and motivation were highly perceived in companies' performance.

• Journal of IKEEE
• /
• v.26 no.3
• /
• pp.468-474
• /
• 2022

In this paper, we propose a method to implement a real-time fire alarm system using deep learning. The deep learning image dataset for fire alarms acquired 1,500 sheets through the Internet. If various images acquired in a daily environment are learned as they are, there is a disadvantage that the learning accuracy is not high. In this paper, we propose a fire image data expansion method to improve learning accuracy. The data augmentation method learned a total of 2,100 sheets by adding 600 pieces of learning data using brightness control, blurring, and flame photo synthesis. The expanded data using the flame image synthesis method had a great influence on the accuracy improvement. A real-time fire detection system is a system that detects fires by applying deep learning to image data and transmits notifications to users. An app was developed to detect fires by analyzing images in real time using a model custom-learned from the YOLO V4 TINY model suitable for the Edge AI system and to inform users of the results. Approximately 10% accuracy improvement can be obtained compared to conventional methods when using the proposed data.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.10a
• /
• pp.188-190
• /
• 2022

We propose how to configure the AI education system based on the purpose of hybrid learning and the teaching-learning principle. Based on the four components of hybrid learning, we have designed the system conceptual diagram and DB configuration diagram for on-line and offline learning environments for effective AI education. The proposed AI education system model in this paper is expected to be a foundation for maximizing the effectiveness of AI education according to the level and needs of learners and building a more effective learner-centered learning environment in cultivating computational thinking in AI education.

• The Journal of the Korea Contents Association
• /
• v.11 no.8
• /
• pp.475-487
• /
• 2011

In this paper, we present a systematic method to derive spectrum of high-order pulse and a novel design of e-Learning system that deals with deriving the spectrum using concept-based branching method. Spectrum of high-order pulse can be derived using conventional methods including 'Consecutive Differentiations' or 'Convolutions', however, their complexity of calculation should be too high to be used as the order of the pulse increase. We develop a recursive algorithm according to the order of pulse, and then derive the formula of spectrum connected to the order with a newly designed look-up table. Moving along, we design an e-Learning content for studying the procedure of deriving high-order pulse spectrum described above. In this authoring, we use the concept-based object branching method including conventional page or title-type branching in sequential playing. We design all four Content-pages divided into 'Modeling', 'Impulse Response and Transfer Function', 'Parameters' and 'Look-up Table' by these conceptual objects. And modules and sub-modules are constructed hierarchically as conceptual elements from the Content-pages. Students can easily approach to the core concepts of the analysis because of the effects of our new teaching method. We offer step-by-step processes of the e-Learning content through unit-based branching scheme for difficult modules and sub-modules in our system. In addition we can offer repetitive learning processes for necessary block of given learning objects. Moreover, this method of constructing content will be considered as an advanced effectiveness of content itself.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.11a
• /
• pp.568-571
• /
• 2020

최근 빅데이터 기술이 다양한 산업과 접목되고 있다. 그 중 고객 이탈 방지가 최우선인 통신사들 또한 예외가 아닐 수 없다. 이에 본 논문은 통신사 데이터에 머신러닝 알고리즘을 접목. 이탈 예측과 데이터 추이를 분석하고, 이를 시각화 하여 일목요연하게 표출하는 과정을 제공함으로서 통신사의 고객 유치 정책을 위한 토대를 마련할 것이다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2022.01a
• /
• pp.93-96
• /
• 2022

본 논문에서는 카메라 센서만을 이용하여 손바닥 사진을 촬영하고 추출된 데이터들을 합성하여 멀티 채널 이미지를 생성 및 분류 모델에 입력하여 신원을 확인하는 딥러닝 모델을 제안한다. 이 모델은 손바닥 사진이 입력되면 손바닥 및 손금 세그멘테이션을 이용하여 마스크 이미지를 추출하고 단일 채널로 구성된 이미지들을 멀티 채널 이미지로 합성/재구성하여 신원을 분류하는 딥러닝 모델이다. 이는 카메라 센서 외 다른 센서가 필요 없다는 장점을 가지고 있으며, 비접촉 신원 인식 시스템에 적용할 수 있다.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.545-548
• /
• 2023

본 논문에서는 골프 코스 시뮬레이션을 위해 수집된 데이터의 정제 및 처리에 요구되는 딥 러닝 모델과 모델 적용 과정에 대해서 논의한다. 최근 스크린 골프 시장의 확대와 골프 시뮬레이터 기술의 발전으로, 위성 이미지, 항공 촬영 이미지, 공간 정보 시스템 (GIS) 등 다양한 데이터 소스로부터 골프 코스에 대한 정보를 수집에 대한 요구가 증가하였다. 이번 연구에서는 이러한 데이터 소스로부터 생성된 원시 데이터를 최적의 시뮬레이션 입력으로 변환하기 위한 컴퓨터 비전 기법과 딥 러닝 모델 구조에 대해서 검토한다. 특히, 데이터에서 골프 코스 시뮬레이션에 요구되는 메타 데이터를 도출하기 위해 코스 분할(Segmentation)과 코스 오브젝트 분류(Classification) 모델을 적용하는 과정을 다룬다. 이를 통해, 본 연구는 골프 코스 시뮬레이터의 개발 과정에서 중요한 기술 요소를 제공하며, 이는 시뮬레이션의 정확도와 골프 코스의 다양성을 증진시키는데에 기여한다.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.425-431
• /
• 2023

음성 어시스턴트 시스템에서 발화의 의도를 분류하고 새로운 의도를 탐지하는 것은 매우 중요한 작업이다. 끊임없이 인입되는 새로운 발화로 인해 기존에 학습된 모델의 의도 분류 성능은 시간이 지남에 따라 점차 낮아진다. 기존 연구들에서 새로운 의도 발견을 위해 제안되었던 클러스터링 방법은 최적의 클러스터 수 결정과 명명에 어려움이 있다. 이러한 제한 사항을 보완하기 위해, 본 연구에서는 대규모 언어 모델 기반의 효과적인 의도 발견 방법을 제안한다. 이 방법은 기존 의도 분류기로 판단하기 어려운 발화에 새로운 의도 레이블을 할당하는 방법이다. 새롭게 인입되는 OOD(Out-of-Domain) 발화 내에서 오분류를 찾아 기존에 정의된 의도를 탐지하고, 새로운 의도를 발견하는 효율적인 프롬프팅 방법도 분석한다. 이를 액티브 러닝 전략과 결합할 경우, 분류 가능한 의도의 개수를 지속 증가시면서도 모델의 성능 하락을 방지할 수 있고, 동시에 새로운 의도 발견을 자동화 할 수 있다.

• Journal of Service Research and Studies
• /
• v.11 no.4
• /
• pp.67-76
• /
• 2021

The importance of smart learning is increasing as the speed of development of non-face-to-face services increases due to the influence of COVID-19. This study is the user satisfaction evaluation model that utilizes the causal relationship between variables used for evaluation, focusing on the usability evaluation results of the learning disability intervention service (www.basic-edu.net) according to the need to evaluate the use satisfaction of the smart learning system. To this end, theoretical studies were conducted on smart learning and learning disability intervention services, www.basic-edu.net, usability evaluation of learning disability intervention systems, and use satisfaction evaluation models. And based on the results, a hypothesis was presented on the user satisfaction evaluation model of the smart learning system. The experimental method allowed 40 students and parents across the country to use the www.basic-edu.net service and was evaluated for its usability. In addition, using this data, the hypothesis was verified using regression analysis based on four variables: ease of use, interest, self-learning, and satisfaction with use. As a result of the hypothesis verification, it was found that the causal relationship of all hypotheses from H1 to H4 was significant.