http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
최지우(Jiwoo Choi),최상일(Sangil Choi),강태원(Taewon Kang) 한국정보기술학회 2022 한국정보기술학회논문지 Vol.20 No.11
Various studies exist to identify individuals. Personal identification research based on inertial data, that is, acceleration and angular velocity acquired with an inertial sensor, is also one of these efforts. In fact, when learning inertial data with CNN, individuals can be identified with high accuracy. However, the individual identification model using inertial data significantly lowers the identification performance when the shoes worn by the individual change. This paper deals with improving this problem by using gait cycle data extracted from inertia data. First, the CNN model using the gait cycle was implemented, and then the model was evaluated using the representative performance evaluation indicators, such as accuracy, precision, recall, and F1-score. As a result, it was confirmed that the proposed model can identify individuals with more than 90% accuracy even when the shoes worn are different.
최지우(Jiwoo Choi),유형진(Hyungjin Yoo),최상일(Sangil Choi),강태원(Taewon Kang) 한국정보기술학회 2021 한국정보기술학회논문지 Vol.19 No.8
In the gait cycle, heel strike (HS) and toe off (TO) of both feet are repeatedly included, and identification of the gait cycle is to find these activities within the cycle. In this paper, we study a method to identify the gait cycle using LSTM, a representative recurrent neural network model. Learning data were collected by performing a gait experiment 20 times each on 50 study participants, 25 males and females each. When performing the experiment, they wear an inertial sensor(acceleration and angular velocity) on their left wrist for collecting gait data. As a result of obtaining the average precision and recall after learning through the LSTM model, the precision was 95.98% and the recall rate was 93.18%. Through this, it can be confirmed that it is an effective method to identify the gait cycle by learning the data obtained from the inertial sensor with the proposed LSTM model.
최지우(Jiwoo Choi),최상일(Sangil Choi),강태원(Taewon Kang) 한국정보기술학회 2022 한국정보기술학회논문지 Vol.20 No.10
In a society centered on hyper-connectivity, as important as information sharing is that each piece of information must be viewed only by legitimate users. In this study, we propose a smartphone authentication system based on human gait, breaking away from the traditional authentication method. After learning human gait with CNN, it is mounted on a smartphone to determine whether the user is a legitimate user by walking for 7 seconds while carrying the smartphone. Accuracy, precision, recall, F1-score, and EER were applied as evaluation indicators of the model proposed in this study. As a result, accuracy, precision, recall, and F1-score all achieved an average of 95% or more, and the average EER was 0.048. What the system analysis results show is that the system proposed in this study has high reliability and low error rate. As a result, this study showed the possibility that human gait could be used as a new user authentication method.
조선 초기 관료의 관청이동을 통해 본 주요 통치기구의 위상 - HAVNet 자료를 중심으로 -
최상일 ( Choi Sangil ),백승민 ( Paek Seungmin ),최지우 ( Choi Jiwoo ),예홍진 ( Yeh Hongjin ),이상국 ( Lee Sangkuk ) 수선사학회 2021 史林 Vol.- No.75
This paper is an introductory study to comprehensively grasp the betweenness centrality of government offices in the early Joseon dynasty using HAVNet data. The process of extracting data around the “Annals of the Joseon Dynasty” and designing and building HAVNet as the basis for research was conducted by history and computer science researchers, which were not the usual method in Korean historical research until now. The approach and analysis results carried out in this study are fundamentally different from previous studies related to the government organizations in the early Joseon dynasty. It is unique study in that it was analyzed based on the contents recorded in the Annals of the Joseon dynasty, not on previous research framework of analysis of state administration and government organizations in the early Joseon dynasty. Based on this study, we will conduct an interdisciplinary research that comprehensively analyzes blood ties and relationships among all historical figures of the Joseon dynasty.
코로나 백신에 정부 정책 발표에 대한 여론 형성에 관한 Causal Impact 분석
구현지(Gu Hyeonji),박채연(Park Chaeyeon),최지우(Choi Jiwoo),노성진(Noh Seongjin) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
본 연구에서는 정부의 코로나 백신 정책이 사람들의 여론에 영향을 얼마나 미쳤는지 알기 위해, 빅카인즈의 뉴스와 유튜브의 댓글 데이터를 수집하여 분석하였다. 먼저 여론의 동향을 알아보기 위해서 Word Cloud 분석을 수행하였고, Causal Impact 분석을 통해 정부 정책이 여론에 얼마나 영향을 분석하였다. 분석 결과 정부의 정책이 뉴스를 통해 통계적으로 유의미한 영향을 미치는 것을 확인하였다. In this study, to find out how much the government"s COVID-19 vaccine policy affected people"s public opinion, BigKinds"s news and YouTube"s comment data were collected and analyzed. First, Word Cloud analysis was performed to find out the trend of public opinion, and how much influence the government policy has on public opinion was analyzed through causal impact analysis. As a result of the analysis, it was confirmed that the government"s policy had a significant through the news statistically.
머신러닝에 의한 기상상태와 COVID-19 감염 관계 분석
이재현(JaeHyun Lee),지승연(SeungYeon Ji),최원준(Wonjun Choi),강태훈(TaeHoon Kang),김상현(SangHyeon Kim),최지우(JiWoo Choi),강태원(TaeWon Kang) 한국정보기술학회 2021 Proceedings of KIIT Conference Vol.2021 No.11
COVID-19 사태의 심각성과 함께 역학조사의 중요성이 시간이 지난 지금까지도 대두되고 있다. 본 연구는 효과적인 역학조사를 위해 활동량에 영향을 끼치는 요인 중 기상정보를 선택하여 확진자의 발생 여부와 기상정보의 상관관계를 머신러닝 모델을 사용하여 분석하고자 한다. 사용한 모델은 SVM(Support Vector Machine) 으로 기본적인 기상정보와 추가적으로 추출한 특성들을 입력으로 다음날의 확진자의 발생 여부를 예측했으며 정확도 81.65%, 정밀도 98%, 재현율 71%의 성능을 보였다. 결과, 유의미한 성능향상을 보였던 대표적인 특성은 일정 기간동안의 확진자 수, 요일이 있었으며 기상정보의 경우 추가적인 특성추출에도 의미있는 성능향상을 보이지 않아 확진자 발생 여부와 기상정보의 관계가 뚜렷하지 않다는 것을 확인할 수 있었다. Along with the seriousness of the COVID-19 incident, the importance of epidemiological investigations is still emerging over time. This study aims to analyze the correlation between the occurrence of confirmed patients and weather information by selecting weather information among the factors influencing the amount of activity for effective epidemiological investigation using a machine learning model. The model used was SVM (Support Vector Machine), which predicted the occurrence of confirmed cases the next day by inputting basic weather information and additional extracted characteristics, and showed performance of 81.65%, precision 98%, and reproduction rate 71%. As a result, it was confirmed that the typical characteristics of significant performance improvement were the number of confirmed cases and days of the week, and in the case of weather information, there was no clear correlation between confirmed cases and weather information.