http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
군집 소실 채널 상에서의 인터리빙된 짧은 코드의 성능 분석
장재윤(Jae-Yoon Jang),장민(Min Jang),김상효(Sang-Hyo Kim),이성준(Sung-Jun Lee) 한국방송·미디어공학회 2009 한국방송공학회 학술발표대회 논문집 Vol.2009 No.11
본 논문에서는 군집 소실 채널 상에서 동작할 수 있는 짧은 길이의 인터리빙 된 코드들의 성능을 분석한다. 먼저 좋은 성능을 갖는 짧은 길이의 그래프 부호와 해밍부호를 설계한다. 이 후 군집 소실에 잘 대응하기 위하여 인터리빙 기능을 채널 부호화 방법에 적용한다. 생성된 짧은 코드에 적용한 인터리빙 부호를 군집 소실에 최적의 성능을 보이는 Reed-Solomon (RS) 부호와 성능을 비교한다. 짧은 길이의 부호이므로, ML(Maximum Likelihood)방법과 BP(Belief propagation)의 두 가지 복호방법들을 이용한 경우 성능의 차이 또한 비교해 본다.
원저(原著) : 유세포 분석기를 이용한 망상적혈구수 측정의 평가
장재윤 ( Jae Yoon Jang ),이현교 ( Hyun Kyo Lee ),오원숙 ( Won Sook Oh ),이경자 ( Kyung Ja Lee ) 대한임상병리사협회 1994 임상혈액검사학회 발표자료집 Vol.1 No.1
Flow cytometric analysis of reticulocyte counts, using membrane-permeable fluorescent dye,thiazole orange, was evaluated as an alternative to the conventional new methylene blue method, and evaluated the referance values of mean channel fluorescence(MCF) in addition to routine reticulocyte count for normal adults. Eighty-eight clinical specimens from inpatients and outpatients in our hospital were analyzed reticulocyte counts by flow cytometer and simultaneously by conventional method. And 40 samples from healthy adults, 20 males and 20 females were analyzed reticulocyte counts and MCF by flow cytometer. We compared reticulocyte counts of new methylene blue method(X) and flow cytometry(Y), the correlation coefficient was 0.90(Y=0.81+0.77) in 88 patient samples(P>0.01). Reference values(mean 2SD) of reticulocyte count were 0.60-2.12 % in adult males(n=20), 0.60-2.23 % in adult females, values of females were slightly higher than those of males. And reference values of MCF were 44.4-66.0 in adult males, 41.3-72.1 in adult females. The flow cytometric fluorescence reticulocyte enumeration methods are efficient, rapid, and reliable for reticulocyte counting.
LPS로 자극된 Raw 264.7 대식세포에서 오미자 씨앗오일의 항염증 효과
장재윤 ( Jae Yoon Jang ),박근혜 ( Geun Hye Park ) 대한본초학회 2015 大韓本草學會誌 Vol.30 No.6
Objectives : This study was designed to investigate of the anti-inflammatory effects of Schisandra chinensis seed oil(SSO) on the production of pro-inflammatory substances in lipopolysaccharide (LPS)-stimulated RAW 264.7 macrophages. Methods : SSO was measured the production of pro-inflammatory factor (NO, PGE2, IL-1β iNOS and, COX-2) in lipopolysaccharide (LPS)-stimulated RAW 264.7 macrophages. we used the following methods : cell viability assay, Griess reagent assay, enzyme-linked immunosorbent assay, Western blotting analysis. Results : The cell viability of SSO(0∼500 μl/mL) processing group was 96.9% and the processing of SSO didn``t have an effect on the cytotoxicity. The inhibitory effect of the nitric oxide (no) production of SSO(500 μg/mL, 50 μg/mL, 10 μg/mL) was each 70.3%, 37.6% and 26.5%. IL-1β production inhibition ability of SSO(500 μg/mL, 100 μg/mL) was each 49.88% and 48.8%. PGE2 production inhibition ability of SSO(500 μg/mL, 100 μg/mL) was each 49.88% and 73.1%, 70.5%. By using SSO, it experimented about iNOS protein expression inhibition ability, that is the NO production enzyme. iNOS protein expression increased in the group processing LPS independently. iNOS protein expression decreased in the group processing SSO together. The expression of the COX-2 protein decreased 89.6%, 81.8% in the group processing SSO. The significance was in the relationship with NO formation inhibition with the relationship with the PGE2 formation inhibition and iNOS protein, it confirmed in SSO with the COX-2 protein. Conclusions : Stimulation of the RAW 264.7 cells with LPS caused an elevated production of nitric oxide (NO), IL-1β and PGE2 which was markedly inhibited by the pretreatment with SSO without causing any cytotoxic effects. The reduced expressions of iNOS protein were consistent with the reductions in NO production in the culture media. SSO may be useful for the treatment of various inflammatory diseases.
다차원 특징 및 가중블록 투표방식을 이용한 얼굴 인식 성능 향상 기법
장재윤(Jae-Yoon Jang),윤호섭(Ho-Sub Yoon) 대한전자공학회 2015 대한전자공학회 학술대회 Vol.2015 No.6
This paper proposes novel block voting measure for improving user face recognition system performance. In order to exploit sample’s global and local characteristic, We use various feature and make large dimension feature by concatenation. After then, compare gallery set with probe image. In this step, we divide feature into some block, and recognize each block using cosine similarity. And we obtain probe image class using weighted voting which calculated by block recognition result. This proposed method has advantage that is robustness for noise or outlier data. Because this method depend on the number of sub-block, if enough sub-blocks are guaranteed, it can make improved performance on face recognition system.
유철환(Cheol-Hwan Yoo),김호원(Ho-Won Kim),한병옥(Byung-Ok Han),장재윤(Jae-Yoon Jang),유장희(Jang-Hee Yoo) 대한전자공학회 2021 전자공학회논문지 Vol.58 No.12
동영상에 포함된 반복적, 주기적 구간을 검출하기 위한 기술은 컴퓨터 비전 분야에서 활발히 연구되고 있다. 기존의 기법들은 일반적으로 반복적 구간 검출을 위한 중간 표현으로서 자기 유사성 행렬(SSM)을 생성하여 활용한다. 그러나 기존의 기법들은 단일 스케일에서의 자기 유사성 행렬의 활용으로 인해 다양한 길이 및 스케일의 반복적 행동을 포함한 동영상에 대해 검출 정확도가 떨어지는 한계점을 갖는다. 이러한 한계점을 극복하기 위해 제안하는 네트워크의 인코더에서는 먼저 3차원 합성곱 신경망의 여러 계층에서 추출된 특징 벡터를 활용하여 다양한 시간적 스케일에 대한 정보를 갖는 자기 유사성 행렬을 생성한다. 이렇게 생성된 자기 유사성 행렬들을 멀티 스케일 특징 앙상블 모듈을 통해 멀티 스케일 U-Net의 입력으로 제공함으로써 동영상 내 다양한 길이의 반복적 구간을 효율적으로 검출한다. 제안하는 기법은 Countix, PERTUBE 데이터셋에서의 실험을 통해 기존의 핸드 크래프트 특징 기반의 기법들뿐만 아니라 딥러닝을 활용한 최신 기법들보다 우수한 검출 성능을 보였다. Recently, techniques for detecting repetitive and periodic segments in a video have been extensively studied in the field of computer vision. Conventional methods typically generate and utilize a self-similarity matrix as an intermediate representation for identifying repetitive segments in a video. However, these methods rely on a single-scale self-similarity matrix(SSM) and thus have a limitation that classification accuracy drops for videos including repetitive segments with various lengths and scales. To solve these problems, the encoder of the proposed network firstly generates self-similarity matrices, which incorporate information on various temporal scales by utilizing feature vectors extracted from multiple layers of the 3D CNN. By providing generated self-similarity matrices as input of a multi-scale U-Net through a multi-scale feature ensemble module, repetitive segments of various lengths in the video can be efficiently detected. Extensive experiments on the Countix and PERTUBE datasets demonstrate that the proposed network not only outperforms most hand-craft feature-based methods but also the latest deep learning-based methods.