건설 근로자 근골격계 부상 방지를 위한 스마트 인솔 기반 들기작업 위험상태 분류 기술

박근형,양강혁 한국건축시공학회 2025 한국건축시공학회지 Vol.25 No.6

근골격계 질환(Work-related musculoskeletal disorders)은 산업재해 중 가장 많은 비중을 차지하는 질환으로, 특히 중량물 을 다루는 작업이 빈번한 건설 분야에서 두드러지게 발생한다. 그러나 건설현장은 작업 위치가 다양하고 여러 작업자가 동 시에 작업을 수행하기 때문에 개별 근로자에게 가해지는 신체부하를 실시간으로 모니터링하기가 쉽지 않다. 이에 본 연구 는 발바닥 압력 데이터를 활용하여 중량물 들기 작업에서의 근골격계 위험 수준을 인식하는 딥러닝 기술을 제안하였다. 데 이터 획득을 위한 실험은 통제된 실험실 환경에서 수행되었으며, 다양한 중량과 작업 상황에서 발바닥 압력 데이터를 수집 하였다. 딥러닝 모델로 양방향 장기단기기억(Bidirectional Long Short-Term Memory)을 적용하였으며, 양발 전체 발바닥 압력 데이터를 사용할 경우 87.7%의 정확도를, 발가락·아치·뒤꿈치 부위 데이터만 사용할 경우 88.2%의 정확도를 달성하 였다. 본 연구 결과는 발바닥 압력 데이터를 통해 들기작업 환경에 따른 위험 상태 인식이 가능하며, 건설근로자의 실시간 근 골격계 위험 평가 및 관리에 유용한 정보를 제공할 수 있음을 보여준다. Work-related musculoskeletal disorders(WMSDs) are the most frequently reported occupational injuries, particularly in construction, where heavy manual handling tasks are common. Monitoring individual workers’ physical loads in such dynamic environments is difficult due to simultaneous tasks across different site locations and limited supervision. This study proposes a smart insole–based approach to classify lifting workload levels using plantar pressure data. Experiments were conducted in a controlled laboratory setting, where plantar pressure data were collected under varying lifting weights and work layouts. A Bidirectional Long Short-Term Memory(Bi-LSTM) model was applied for classification. The model achieved an accuracy of 87.7% when using full plantar pressure data from both feet and a slightly higher accuracy of 88.2% when using pressure data from the toe, arch, and heel regions. The results demonstrate that plantar pressure data can provide reliable information for evaluating and managing lifting workloads in construction environment.

보철용 크라운/어버트먼트 갭이 표면적합도에 미치는 영향

박근형 대한치과기공학회 2007 대한치과기공학회지 Vol.29 No.1

차의 기능성 성분과 생리활성

박근형,문제학 한국차학회 2006 한국차학회지 Vol.12 No.1

차의 성분과 기능성에 관한 기능성분과 생리활성, 차의 향기

단일층 다결정 실리콘 플래시 EEPROM 셀의 설계, 제작 및 전기적인 특성 분석

박근형 충북대학교 컴퓨터정보통신연구소 2012 컴퓨터정보통신연구 Vol.20 No.1

In this paper, a new single polysilicon flash EEPROM's cell with a select transistor was proposed. The device was designed and fabricated using the Hynix 0.35 μm CMOS processes, and then the electrical characteristics of the fabricated devices were measured. From the measurement results, it was found that the devices showed very excellent performance of the program, the erase and the read operations. In addition, the devices showed very excellent endurance characteristics. Therefore it is sure that the single polysilicon flash EEPROM's cell proposed in this paper can be adequate for an embedded EEPROM device in the future.

국산 녹차의 생산 및 산업동향

박근형 한국식품과학회 2008 국제녹차심포지엄 Vol.2008 No.1

우리나라 스태그플레이션 발생확률 추정

박근형,강규호 한국은행 2022 經濟分析 Vol.28 No.3

The prediction of the short- and long-term stagflation probability has a profound influence on the Bank of Korea's preemptive monetary policy decision-making for economic and price stability. This study estimates the probability of stagflation in Korea over the next two years. To this end, a joint predictive distribution of the inflation rate and the real economic growth rate is generated using univariate and bivariate autoregressive distributed lag (ADL) models. In this process, the optimal prediction model is selected through Bayesian variable selection and precise tuning through out-of-sample prediction. As a result of long-term and short-term out-of-sample predictions for the last 5 years, the joint distribution predictive accuracy was maximized when each variable was independently predicted using the univariate ADL model in most forecasting horizons. As a result of the actual prediction using the optimal model, the probability of stagflation (GDP growth rate below 1% and CPI inflation rate above 4% or two consecutive quarters of decline in real GDP level and CPI inflation rate above 4%) temporarily increased in the fourth quarter of 2022 and then dropped significantly to less than 10%. From these results, it is predicted that the possibility of stagflation is very limited. 장단기 스태그플레이션 발생확률 예측은 한국은행의 경기 및 물가안정을 위한 선제적 통화정책 의사결정에 지대한 영향을 미친다. 본 연구는 향후 2년간 우리나라 스태그플레이션 발생확률을 추정한다. 이를 위해 일변수 및 이변수 자기회귀시차(Autoregressive Distributed Lag, ADL) 모형을 이용하여 물가상승률과 경제성장률의 결합예측분포를 생성한다. 이 과정에서 표본외 예측을 통해 베이지안 변수선택과 정밀한 튜닝과정을 거쳐 최적 예측모형을 선택한다. 최근 5년을 대상으로 한 장단기 표본외 예측결과, 대부분의 예측시계에서 일변수 ADL 모형으로 각 변수를 독립적으로 예측했을 때 결합분포 예측력이 극대화되었다. 최적 모형을 이용한 실제 예측결과, 스태그플레이션(GDP 성장률 전년 동기비 1% 이하와 소비자물가 상승률 4% 이상, 또는 2분기 연속 전기비 마이너스 성장과 물가상승률 4% 이상) 발생확률은 금년 4분기 중 일시적으로 높아지나 이후 10% 이하 수준으로 크게 하락하는 것으로 나타났다. 이러한 결과로 볼 때 향후 스태그플레이션이 발생할 가능성은 매우 제한적인 것으로 추정된다.

건설 근로자의 족저압 정보를 활용한 신체부하 인식 기술

박근형,황영서,고성석,양강혁 대한건축학회지회연합회 2023 대한건축학회연합논문집 Vol.25 No.5

A lot of construction works are done by manually and often involves heavy materials handling, which increases the risk of musculoskeletal disorders. However, monitoring physical load levels applied to workers during construction work is difficult due to the large size of the site and a huge number of workforce. Under this circumstance, this study developed an approach to evaluate the lifting workload of construction workers using a smart insole sensor for the purpose of preventing musculoskeletal disorders. In the experiment, different level of risks were set by changing the lifting load according to the NIOSH Lifting Index. Participants wore a smart insole and performed repetitive lifting tasks. A analysis was conducted by applying the Bi-LSTM model, a deep learning algorithm based on a recurrent neural network. As a result of the analysis, an accuracy of up to 84.1% was confirmed when using data collected from the nearest foot to the lifting object. The approach introduced in this study utilizes foot-pressure data which is easier to acquire than other biometric data and would have a higher field applicability. The approach would help to manage the level of physical load during a heavy material handling tasks at construction sites and prevent musculoskeletal disorders of construction workers.

L2 learner awareness of Negative Polar Questions and potential/limitation of instruction

박근형 언어과학회 2023 언어과학연구 Vol.- No.107

This study examines English and Korean NPQs in L2 learning, focusing on the challenges faced by Korean learners in understanding English NPQs, particularly those with low negation. Conventional teaching methods emphasizing a dichotomy between answering patterns hinder natural acquisition. The study recommends balanced examples of NPQ types in textbooks, breaking the notion that NPQs mirror PPQs and emphasizing contextual biases. Understanding cross-linguistic similarities and differences in NPQs is crucial for mastering their complexities. Recognizing the cultural contexts in which these questions are used is important in grasping their nuanced interpretations. This comprehensive approach aims to aid L2 learners in interpreting NPQs effectively in classrooms.

드론 영상을 통한 도로상태 정보수집 기술 개발

박근형 한국ITS학회 2019 한국ITS학회 학술대회 Vol.2019 No.11

다중 채널 동적 객체 정보 추정을 통한 특징점 기반 Visual SLAM

박근형,조형기 대한임베디드공학회 2024 대한임베디드공학회논문지 Vol.19 No.1

An indirect visual SLAM takes raw image data and exploits geometric information such as key-points and line edges. Due to various environmental changes, SLAM performance may decrease. The main problem is caused by dynamic objects especially in highly crowded environments. In this paper, we propose a robust feature-based visual SLAM, building on ORB-SLAM, via multi-channel dynamic objects estimation. An optical flow and deep learning-based object detection algorithm each estimate different types of dynamic object information. Proposed method incorporates two dynamic object information and creates multi-channel dynamic masks. In this method, information on actually moving dynamic objects and potential dynamic objects can be obtained. Finally, dynamic objects included in the masks are removed in feature extraction part. As a results, proposed method can obtain more precise camera poses. The superiority of our ORB-SLAM was verified to compared with conventional ORB-SLAM by the experiment using KITTI odometry dataset.