Particle Swarm Optimization Using Adaptive Boundary Correction for Human Activity Recognition

( Yongjin Kwon ),( Seonguk Heo ),( Kyuchang Kang ),( Changseok Bae ) 한국인터넷정보학회 2014 KSII Transactions on Internet and Information Syst Vol.8 No.6

As a kind of personal lifelog data, activity data have been considered as one of the most compelling information to understand the user`s habits and to calibrate diagnoses. In this paper, we proposed a robust algorithm to sampling rates for human activity recognition, which identifies a user`s activity using accelerations from a triaxial accelerometer in a smartphone. Although a high sampling rate is required for high accuracy, it is not desirable for actual smartphone usage, battery consumption, or storage occupancy. Activity recognitions with well-known algorithms, including MLP, C4.5, or SVM, suffer from a loss of accuracy when a sampling rate of accelerometers decreases. Thus, we start from particle swarm optimization (PSO), which has relatively better tolerance to declines in sampling rates, and we propose PSO with an adaptive boundary correction (ABC) approach. PSO with ABC is tolerant of various sampling rate in that it identifies all data by adjusting the classification boundaries of each activity. The experimental results show that PSO with ABC has better tolerance to changes of sampling rates of an accelerometer than PSO without ABC and other methods. In particular, PSO with ABC is 6%, 25%, and 35% better than PSO without ABC for sitting, standing, and walking, respectively, at a sampling period of 32 seconds. PSO with ABC is the only algorithm that guarantees at least 80% accuracy for every activity at a sampling period of smaller than or equal to 8 seconds.

이동 객체 위치 추정을 위한 저조도 환경에서의 특징점 매칭 방법론

권용진(Yongjin Kwon),윤준희(Junhee Youn),최강혁(Kanghyuk Choi) 한국산학기술학회 2023 한국산학기술학회논문지 Vol.24 No.10

최근 스마트시티, 도시계획, 로봇 공학, 공간정보 등 다양한 분야에서 이동 객체의 위치 추정 기술은 핵심 알고리즘으로 사용될 수 있다는 점에서 다양한 연구가 진행되고 있다. 영상 기반의 위치추정 방법은 비교적 구현이 간편하고 안정적이지만, 저조도 환경에서는 대비 저하와 노이즈로 인해 객체의 특징점을 정확하게 추출 및 매칭 하는 것에 한계가 존재한다. 따라서 본 논문에서는 저조도 환경에서 영상 기반의 위치추정을 위한 특징점 매칭 성능을 향상시키는 방법론을 제안하였다. 본 연구에서는 대표적인 특징점 매칭 알고리즘 KAZE, ORB, SIFT 그리고 SURF를 이용하여 저조도 환경에서 특징점 매칭을 수행하고 그 결과에 대해 비교 분석한다. 다음으로 CLAHE, 미디안 필터 그리고 다운 샘플링의 영상 전처리 기법을 이용하여 특징점 매칭 알고리즘과 함께 조합하여 성능을 비교 분석하였다. 분석 결과 ORB 알고리즘과 다운 샘플링의 조합이 평균 92% 이상의 정확도와 최대 0.287초 이하의 처리 시간을 보였다. 이는 다른 전처리 기법 및 특징점 매칭 알고리즘 조합 대비 가장 높은 정확도와 함께 우수한 성능을 보여준다. Technology for estimating the position of moving objects has been the subject of extensive research due to its potential in diverse fields, such as smart cities, urban planning, robotics, and spatial information. Image-based location-estimation methods are relatively simple to implement and stable. However, in low-light environments, there are limitations in accurately extracting and matching feature points of objects due to reduced contrast and noise. This paper presents a framework to enhance feature-matching performance for image-based position estimation in low-light environments. We employed well-known feature matching algorithms such as KAZE, ORB, SIFT, and SURF to perform feature-point matching in low-light environments and analyzed the results. We then combined these algorithms with image-processing techniques like CLAHE, a median filter, and down-sampling to compare their performance. The combination of the ORB algorithm and down-sampling showed an average accuracy of more than 92% and a maximum processing time of less than 0.287 seconds. This combination showed excellent performance with the highest accuracy compared to other combinations of image-processing techniques and feature-point matching algorithms.

컴퓨팅 서버 에너지 효율화를 위한 워크로드 환경 별 팬 제어 알고리즘 비교

권용진(YongJin Kwon),안재훈(JaeHoon An),김영환(YoungHwan Kim) 대한전자공학회 2024 대한전자공학회 학술대회 Vol.2024 No.6

As the demand for computing server systems such as artificial intelligence, big data, and cloud increases, energy consumption is also rapidly increasing. In particular, energy used for cooling accounts for the second largest proportion of total energy consumption following ICT infrastructure equipment such as servers and storage devices. Additionally, many servers remain idle, consuming power but providing no service. In order to save power, various modeling techniques are being researched to track and predict data center power consumption, and research is being conducted to realize power savings through data center structural efficiency or cooling system technology. However, there is a lack of comparative research on under what circumstances each method is effective. In this paper, we compare and analyze the efficiency of several fan control algorithms for each workload type. We discuss how to implement different fan control algorithms that can achieve server power savings in idle and operational states, and analyze the effectiveness of each algorithm by workload type. Through this, we discuss ways to save energy in computing server cooling systems through optimal fan control according to workload.

건강보험 청구자료를 활용한 약물관련 악골괴사의 국내 현황 및 비용 분석

권용진(Yongjin Kwon),박미혜(Mi-Hai Park) 대한약학회 2020 약학회지 Vol.64 No.4

Although Medication-Related Osteonecrosis of the Jaw (MRONJ) does not usually occur, the use of drugs that may cause MRONJ is increasing and there is no research reported about its cost and the perception of MRONJ is low in Korea. So here we analyzed the current disease status, such as occurrence and duration, and the cost of treatment of MRONJ in Korea. It was analyzed by the National Health Insurance Service sample cohort 2.0DB data. Patients were included who diagnosed and treated with ‘Osteonecrosis due to drugs (ICD-10: M87.1)’ or ‘Inflammatory conditions of jaws (ICD-10: K10.2)’ after the less than three years from the last dose of drugs. A total of 183 MRONJ patients were identified. The most frequently administered Bone-Modifying Agents were Alendronate, Risedronate, and Ibandronate. From 2002 to 2015, the proportion of MRONJ in the total number of prescribed patients was between 0.2% and 0.9%. The billing costs for MRONJ per patient was KRW 934,422. The highest cost claim was Surgery of Osteomyelitis of Mandible or Maxilla. The total cost of medical, non-medical and indirect costs was KRW 1,891,808 per patient. Through this research, it was found that the cost of treatment of MRONJ was high, and it was possible to analyze the current status and characteristics of MRONJ in Korea.

실제 의료 데이터 분석을 통한 데이터 익명화 방법 제시

권용진(Yongjin Kwon),연종흠(Jongheum Yeon),이상구(Sang-goo Lee) 한국정보과학회 2011 한국정보과학회 학술발표논문집 Vol.38 No.1C

정부, 병원, 공공기관, 기업 등에서 많은 양의 개인 정보를 수집하고, 다양한 목적으로 수집한 데이터를 공개하기도 한다. 프라이버시 보호를 위해 공개할 데이터를 어떻게 익명화를 할 것인지 많이 연구가 되었지만, 알고리즘 적용의 어려움과 데이터에 대한 비현실적인 가정 때문에 실제 데이터에 적용되는 사례는 찾아보기 힘들다. 본 논문에서는 먼저 다양한 익명화 방법에 대한 분류를 하고, 장단점을 살펴본다. 그리고 의료 기관에서 의료 데이터를 공개하는 경우를 고려하여 실제 의료 데이터가 갖는 특징을 파악하고, 의료 데이터에 적용할 수 있는 프라이버시 모델(privacy model)과 알고리즘에 적용하기 위한 의료 데이터의 기본 형태에 대해 제시한다.

두 개의 수신안테나를 사용한 채널적응형 임의접근의 포착성능 시뮬레이션

권용진(Yongjin Kwon),문희찬(Hichan Moon) 한국통신학회 2013 한국통신학회 학술대회논문집 Vol.2013 No.11

Deep learning-based scalable and robust channel estimator for wireless cellular networks

Anseok Lee(Anseok Lee),Yongjin Kwon(Yongjin Kwon),Hanjun Park(Hanjun Park),Heesoo Lee(Heesoo Lee) 한국전자통신연구원 2022 ETRI Journal Vol.44 No.6

In this paper, we present a two-stage scalable channel estimator (TSCE), a deep learning (DL)-based scalable, and robust channel estimator for wireless cellular networks, which is made up of two DL networks to efficiently support different resource allocation sizes and reference signal configurations. Both networks use the transformer, one of cutting-edge neural network architecture, as a backbone for accurate estimation. For computation-efficient global feature extractions, we propose using window and window averaging-based self-attentions. Our results show that TSCE learns wireless propagation channels correctly and outperforms both traditional estimators and baseline DL-based estimators. Additionally, scalability and robustness evaluations are performed, revealing that TSCE is more robust in various environments than the baseline DL-based estimators.

The Spillover Effects of Health Information Technology Implementation on Healthcare Quality

Yongjin Park,Youngsok Bang,Juhee Kwon 한국경영정보학회 2018 한국경영정보학회 학술대회논문집 Vol.2018 No.11

U.S. government subsidies under the HITECH Act of 2009 have boosted hospitals’ IT investments, which are expected to improve the quality of care as well as the effectiveness of healthcare management. Given the rush to adopt health information technology (HIT) throughout the continuum of care across healthcare providers, this study tries to identify the spillover effects of HIT adoption on quality of care. Using 1,965 U.S. hospital data in 232 health referral regions (HRRs), we examine how a hospital’s and its neighboring hospitals’ HIT adoptions interact with each other and how they impact readmission rates. We find that a hospital’s readmission rate is reduced by both its own and neighbors’ HIT adoption. Such effects become greater along with the focal hospital’s own adoption. We further investigate how spillover effects vary with HRRs’ different market structures and hospitals’ meaningful-use status. Our findings offer theoretical and managerial insights for both healthcare researchers and practitioners.

Fabrication of a biofuel cell improved by the π-conjugated electron pathway effect induced from a new enzyme catalyst employing terephthalaldehyde

Chung, Yongjin,Hyun, Kyu Hwan,Kwon, Yongchai The Royal Society of Chemistry 2016 Nanoscale Vol.8 No.2

<▼1><▼1><P>A model explaining π-conjugated electron pathway effect induced by enzyme catalyst adopting terephthalaldehyde is suggested with application as enzymatic biofuel cell.</P></▼1><▼2><P>A model explaining the π-conjugated electron pathway effect induced by a novel cross-linker adopted enzyme catalyst is suggested and the performance and stability of an enzymatic biofuel cell (EBC) adopting the new catalyst are evaluated. For this purpose, new terephthalaldehyde (TPA) and conventional glutaraldehyde (GA) cross-linkers are adopted on a glucose oxidase (GOx), polyethyleneimine (PEI) and carbon nanotube (CNT)(GOx/PEI/CNT) structure. GOx/PEI/CNT cross-linked by TPA (TPA/[GOx/PEI/CNT]) results in a superior EBC performance and stability to other catalysts. It is attributed to the π bonds conjugated between the aldehyde of TPA and amine of the GOx/PEI molecules. By π conjugation, electrons bonded with carbon and nitrogen are delocalized, promoting the electron transfer and catalytic activity with an excellent EBC performance. The maximum power density (MPD) of an EBC adopting TPA/[GOx/PEI/CNT] (0.66 mW cm<SUP>−2</SUP>) is far better than that of the other EBCs (the MPD of EBC adopting GOx/PEI/CNT is 0.40 mW cm<SUP>−2</SUP>). Regarding stability, the covalent bonding formed between TPA and GOx/PEI plays a critical role in preventing the denaturation of GOx molecules, leading to an excellent stability. By repeated measurements of the catalytic activity, TPA/[GOx/PEI/CNT] maintains its activity to 92% of its initial value even after five weeks.</P></▼2></▼1>

Sulfenic Acid Doped Mesocellular Carbon Foam as Powerful Catalyst for Activation of V(II)/V(III) Reaction in Vanadium Redox Flow Battery

Chung, Yongjin,Jeong, Jooyoung,Pham, Hien Thi Thu,Lee, Jinwoo,Kwon, Yongchai The Electrochemical Society 2018 Journal of the Electrochemical Society Vol.165 No.11

<P>Sulfur-based functional group doped mesoporous carbon (MC) catalysts are prepared to investigate their effects on V(II)/V(III) and V(IV)/V(V) redox reactions as well as on the performance of vanadium redox flow battery (VRFB). According to the electrochemical analysis, although all of the catalysts applied have a similar influence on the V(IV)/V(V) reaction, for the V(II)/V(III) reaction, the sulfenic acid-doped MC shows better catalytic activity with a faster electron transfer rate and narrower peak potential separation. Nyquist plots also support the superiority of the sulfenic acid-doped MC catalyst with a smaller charge transfer resistance. When the catalysts are loaded onto a negative electrode for measuring the performance of VRFB, the VRFB using sulfenic acid-doped MC catalyst shows 6 similar to 10% higher voltage efficiency and 5 similar to 12% higher energy efficiency than other VRFBs. The excellent catalytic activity of the catalyst for the V(II)/V(III) reaction and superior performance of VRFB are due to an increase in the-SOxH bond that acts as an active site for the reaction. (C) 2018 The Electrochemical Society.</P>