基於深度學習進行模擬健身房單一器材與整體環境之動作辨識

Motion recognition based on deep learning to simulate single equipment and overall environment of gym

陳正鑫(Zheng-Xin Chen) ； 陳五洲(Wu-Chou Chen) ； 何金山(Chin-Shan Ho) ； 李仁軍(Jen-Chun Lee)

《華人運動生物力學期刊》 19卷1期 (2022/03) Pp. 9-14

https://doi.org/10.53106/207332672022031901002

摘要

科技日新月異人們生活品質逐漸在改變，人們從以前書信往來到現在的視訊電話。人類所聽到的聲音、所感受到的溫度及看到的畫面，都是受器接收後，經由神經傳導到大腦，再藉由大腦的綜整處裡所得到的資訊，對於電腦而言，這些聲音、訊息及畫面就是大量的類比及數位訊號組合而成的資訊。如何透過電腦進行影像監控，並將得來的影像進行解析，獲取我們所想要的結果，始終是一項熱門的課題。以目前的運動監測系統來說，現今許多人為了身體健康，會上健身房運動，而戴上運動手環進行監測目前生理機能如何，再配合巡場的人員觀察與指導，以達到運動健身的效果。本研究主要探討如何針對鏡頭攝影的每個人狀態進行標示。若有突發狀況則會提供警示，不再只是單純的監視畫面，而是能真正掌握每個人的狀態監測系統。以研究面向來說，R-CNN、Fast R-CNN、Faster R-CNN及Mask R-CNN等許多以卷積類神經網路Convolutional Neural Networks（CNN）為主軸核心擷取圖像特徵，以達成物件辨識與追蹤技術，其成果已應用在各個領域中。所以本研究藉由羅技C310攝影機，並透過YOLO v4深度學習模型結合即時影像進行人體狀態辨識，配合使用者讓使用者能依據自身需求建立屬於自己的人體狀態資料庫，以達到人體狀態監測的效果。

關鍵字

機械學習 ； YOLO ； 狀態辨識 ； 影像處理

並列摘要

The quality of people's life is gradually changing with the rapid advances in technology, from the old days of correspondence to the current video phone. The sound we hear, the temperature we feel, and the images we see are all received by the receptors, transmitted to the brain through the nerves, and then integrated by the brain to obtain the information. It is always a hot topic to monitor the image through computer and analyze the image to get the result we want. In terms of the current exercise monitoring system, many people nowadays will go to the gym to exercise for their health, and wear exercise bracelets to monitor the current physiological function, and then with the observation and guidance of the patrolling staff, in order to achieve the effect of exercise and fitness. This study focuses on how to mark the status of each person for camera photography. If there is a sudden situation will provide a warning, no longer just monitor the screen, but can really grasp the status of each person monitoring system. In terms of research, R-CNN, Fast R-CNN, Faster R-CNN, Mask R-CNN, and many other Convolutional Neural Networks (CNN) are used as the main axis to capture image features to achieve object recognition and tracking technology, and their results have been applied in various fields. Therefore, this study uses Logitech C310 camera and YOLO v4 deep learning model combined with real-time images for human body status recognition, and allows users to build their own human body status database according to their needs to achieve the effect of human body status monitoring.

並列關鍵字

Mechanical learning ； YOLO ； Status recognition ； Image processing

參考文獻

Bochkovskiy, A., Wang, C.-Y., & Liao, H.-Y. M. (2020). Yolov4: Optimal speed and accuracy of object detection. arXiv preprint arXiv:2004.10934.

Cao, Z., Hidalgo, G., Simon, T., Wei, S.-E., & Sheikh, Y. (2019). OpenPose: realtime multi-person 2D pose estimation using Part Affinity Fields. IEEE transactions on pattern analysis and machine intelligence, 43(1), 172-186.

Cao, Z., Simon, T., Wei, S.-E., & Sheikh, Y. (2017). Realtime multi-person 2d pose estimation using part affinity fields. Paper presented at the Proceedings of the IEEE conference on computer vision and pattern recognition.

Girshick, R. (2015). Fast r-cnn. Paper presented at the Proceedings of the IEEE international conference on computer vision.

Jiang, Z., Zhao, L., Li, S., & Jia, Y. (2020). Real-time object detection method based on improved YOLOv4-tiny. arXiv preprint arXiv:2011.04244.