http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
이정한(Jeonghan Lee),박한훈(Hanhoon Park) 한국통신학회 2022 한국통신학회 학술대회논문집 Vol.2022 No.2
본 논문은 딥페이크(deepfake) 영상 검출을 위해 영상을 패치로 나누고 패치별로 위조 여부를 판별한 후, 각 패치에 대한 판별 결과를 취합하는 방법을 제시하고, StyleGAN2로 생성된 영상에 대한 검출 성능을 실험을 통해 검증한다. 실험 결과, 패치 크기에 따라 검출 정확도는 달라졌으나, 전반적으로 패치 기반 방법은 검출 정확도를 크게 개선할 수 있으며, 신뢰도가 높은 패치를 선별하는 과정을 추가함으로써 검출 정확도를 보다 향상시킬 수 있음을 확인하였다.
이정한(Jeonghan Lee),조경석(Kyungseok Cho),선효성(Hyosung Sun),신형기(Hyungki Shin),이수갑(Soogab Lee) 한국유체기계학회 1998 유체기계 연구개발 발표회 논문집 Vol.- No.-
Aerodynamic noise generated by automobile cooling fan is investigated. Automobile cooling fans radiate both discrete frequency noise as well as broadband noise. In the present work, the former is considered through free-wake panel method coupled with acoustic analogy fully considering the retarded time variation on the blade surface, while the latter is taken into account by three well-established broadband noise components. Experiments were performed to supplement necessary inputs as well as to provide the final comparison with the predicted noise spectrum. The predicted noise levels at blade passing frequencies agree well with the experimental data for the first few harmonics. Although the predicted broadband noise levels at higher frequencies fall below the experimental data due to the fundamental shortcomings of the utilized formulations, the analysis offers a detailed physical understanding of the fan noise generation processes.
연합 학습 기반 분산 FMCW MIMO Radar를 활용한 모션 인식 알고리즘 개발 및 성능 분석
강종성,이승호,이정한,양윤지,박재현,Kang, Jong-Sung,Lee, Seung-Ho,Lee, Jeonghan,Yang, YunJi,Park, Jaehyun 대한임베디드공학회 2022 대한임베디드공학회논문지 Vol.17 No.3
In this paper, we implement a distributed FMCW MIMO radar system to obtain Micro Doppler signatures of target motions. In addition, we also develop federated learning based motion recognition algorithm based on the Micro-Doppler radar signature collected by the implemented FMCW MIMO radar system. Through the experiment, we have verified that the proposed federated learning based algorithm can improve the motion recognition accuracy up to 90%.