http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
하정수,한유정,Ha, Jung-Soo,Han, Yoo-Jeong 한국재료학회 2019 한국재료학회지 Vol.29 No.2
A powder mixture of 70 wt% $Al_2O_3$ and 30 wt% hydroxyapatite (HA) is sintered at $1300^{\circ}C$ or $1350^{\circ}C$ for 2 h at normal pressure. An $MgF_2$-added composition to make HA into fluorapatite (FA) is also prepared for comparison. The samples without $MgF_2$ show ${\alpha}$ & ${\beta}$-tricalcium phosphates (TCPs) and $Al_2O_3$ phases with no HA at either of the sintering temperatures. In the case of $1,350^{\circ}C$, a $CaAl_4O_7$ phase is also found. Densification values are 69 and 78 %, and strengths are 156 and 104 MPa for 1,300 and $1,350^{\circ}C$, respectively. Because the decomposition of HA produces a $H_2O$ vapor, fewer large pores of $5-6{\mu}m$ form at $1,300^{\circ}C$. The $MgF_2$-added samples show FA and $Al_2O_3$ phases with no TCP. Densification values are 79 and 87 %, and strengths are 104 and 143 MPa for 1,300 and $1,350^{\circ}C$, respectively. No large pores are observed, and the grain size of FA ($1-2{\mu}m$) is bigger than that of TCP ($0.7{\mu}m{\geq}$) in the samples without $MgF_2$. The resulting $TCP/Al_2O_3$ and $FA/Al_2O_3$ composites fabricated in situ exhibit strengths 6-10 times higher than monolithic TCP and HA.
Multi-site Split Learning to Preserve Privacy of Patient Data
Yoo Jeong Ha(하유정),Joongheon Kim(김중헌) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.11
With the advancement of the Internet, many firms have utilized this powerful tool to store most of their data online; and hospitals are no exception. Information collected at the hospitals is one of the most private and personal information on an individual. The benefit of sharing medical data between institutions regarding the patient"s health aspect is largely agreed upon. Yet, the issue of information leakage highlights the flaw in electronic health records (EHR)s. Therefore, to overcome the issue of exposing personal information, this paper presents spatio temporal split learning (STSL). This STSL method is a distributed deep neural network that allows different participants to collectively take part in training a deep neural network while protecting the privacy of these sensitive data.
주행 시 행동 인식을 위한 딥러닝기반 비선형 전처리 기법
백한결(Hankyul Baek),하유정(Yoo Jeong Ha),유민재(Minjae Yoo),정소이(Soyi Jung),김중헌(Joongheon Kim) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.6
본 논문은 CNN을 활용한 두가지 딥러닝 모델인 U-Net과 EfficientNet을 활용하여 mmWave Radar를 통하여 측정된 Doppler Range map 이미지 안의 노이즈를 제거 및 동작을 인식하는 방법을 제시하였다. 또한 이를 통하여 다양한 설정 환경에서의 mmWave레이더 데이터 통합 전처리 방식을 제안한다.