Random Forest를 이용한 자세 분류 알고리즘
A Posture Classification Algorithm Using the Random Forest
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
많은 사람들이 컴퓨터 사용 시간이 늘어나면서 잘못된 앉은 자세로 인하여 목과 허리의 질환을 겪는 사람들이 많아졌다. 이러한 질환들을 예방하기 위하여 영상 촬영으로 자세를 분류하는 연구, 신체 혹은 의류에 센서를 부착하여 자세를 측정하는 연구 등 많은 방법들이 제안되었다. 하지만 기존 연구들에서는 데이터 취득 환경, 높지 않은 정확도, 다양하지 않은 자세 분류 등 여전히 한계가 있다. 이러한 한계점들을 보완하기 위하여 본 연구에서는 현대인들의 질환에 악영향을 끼치는 거북목 자세, 뒤로 기댄 자세, 왼쪽으로 기운 자세, 오른쪽으로 기운 자세와 바른 자세를 분류하는 방법을 제안한다. 제안 방법에서는 정면에서 영상을 촬영하고 얼굴과 어깨를 검출하여 좌표와 각도를 계산한다. 그리고 5가지 변수들의 중요도를 측정하여 변수들에 가중치를 부여하고, 이 변수들을 랜덤포레스트에 적용하여 자세들을 분류한다. 실험 결과, 제안 방법에서 98%의 정확도를 보였으며, 기존의 영상 자료를 이용한 연구들과의 비교 실험에서도 최적의 성능을 보였다.
Abstract
As computer usage time increases, many people experience neck and back ailments due to poor sitting posture. To prevent these ailments, various methods have been proposed, including research on classifying posture through video recordings and measuring posture by attaching sensors to the body or clothing. However, existing studies still have limitations such as the data acquisition environment, low accuracy, and limited posture classification. To address these limitations, this study proposes a method to classify postures that negatively affect the health of modern individuals, including forward head posture, reclining posture, leaning to the left, leaning to the right, and correct posture. In the proposed method, videos are recorded from the front, and the face and shoulders are detected to calculate coordinates and angles. The importance of five variables is then measured to assign weights to these variables, and they are applied to a random forest algorithm to classify postures. Experimental results showed that the proposed method achieved 98% accuracy and demonstrated optimal performance compared to existing studies using video data.
참고문헌 (13)
- E. C. P. Chu, "Preventing the progression of text neck in a young man : A case report", Radiology Case Reports, 17(3), pp.978~982 (2022)
- J. Lee, "Classifying Poor Postures of the Neck and Spine in Computer Work by Using Image and Skeleton Analysis", Applied Sciences, 13(19), pp.10935~ (2023)
- K. H. Choi, "A comparison study of posture and fatigue of neck according to monitor types (moving and fixed monitor) by using flexion relaxation phenomenon (FRP) and craniovertebral angle (CVA)", International Journal of Environmental Research and Public Health, 17(17), pp.6345~ (2020)
- T. Toda, "Close face-distance warning system for straightend neck prevention", IECON 2015-41st Annual Conference of the IEEE Industrial Electronics Society, pp.9~12 (2015)
- F. Ji-sheng, "Sitting Posture Detection System Based on Stacking Ensemble Learning", 2022 IEEE 2nd International Conference on Computer Systems (ICCS), pp.23~25 (2022)
- Y. Kuang, "Learner posture recognition via a fusing model based on improved SILTP and LDP", Multimedia Tools and Applications, 78, pp.30443~30456 (2019)
- P. Tokas, "Machine learning based text neck syndrome detection using Microsoft Kinect sensor", Materials Today: Proceedings, 80, pp.3751~3756 (2023)
- H. Y. Tang, "Upper body posture recognition using inertial sensors and recurrent neural networks", Applied Sciences, 11(24), pp.12101~ (2021)
- A. Petropoulos, "SPoMo: IMU-based real-time sitting posture monitoring", 2017 IEEE 7th International Conference on Consumer Electronics-Berlin, pp.3~ (2017)
- Q. Wan, "Hip positioning and sitting posture recognition based on human sitting pressure image", Sensors, 21(2), pp.426~ (2021)
- D. Bibbo, "A sitting posture monitoring instrument to assess different levels of cognitive engagement", Sensors, 19, pp.455~ (2019)
- Breiman, L, "Random forests", Machine learning, 45(3), pp.5~32 (2001)
- A. K. Singh, "Real-time human pose detection and recognition using mediapipe", International Conference on Soft Computing and Signal Processing, pp.18~19 (2021)