簡易檢索 / 詳目顯示

研究生: 李昱勳
Lee, Yu-Hsun
論文名稱: 基於壓力感測器與深度學習之居家運動分析系統
Home Exercise Analysis System Based on Pressure Sensor and Deep Learning
指導教授: 賀耀華
Ho, Yao-Hua
口試委員: 陳伶志
Chen, Ling-Jyh
黃致豪
Huang, Jhih-Hao
賀耀華
Ho, Yao-Hua
口試日期: 2022/12/12
學位類別: 碩士
Master
系所名稱: 資訊工程學系
Department of Computer Science and Information Engineering
論文出版年: 2023
畢業學年度: 111
語文別: 中文
論文頁數: 64
中文關鍵詞: 深度學習CNN運動科學壓力感測器
英文關鍵詞: Deep Learning, CNN, Sports Science, Pressure sensing
研究方法: 實驗設計法
DOI URL: http://doi.org/10.6345/NTNU202300063
論文種類: 學術論文
相關次數: 點閱:151下載:0
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報
  • 2019年新型冠狀病毒的爆發,因其強大的傳播力,人們被迫待 在家中減少外出。運動中心與健身房等場所也容易因為汗水或是 近距離的接觸進而增加染疫的風險。然而大部分的使用者在沒有 教練的協助下,常常導致動作不夠標準或是缺乏系統性的訓練, 造成運動傷害以及成效有限。因此本研究提出 Home Exercise Analysis System Based on Pressure Sensor and Deep Learning,簡稱 為 HomeXDL。HomeXDL 運用壓力及壓力重心變化結合深度學習 來達到運動紀錄以及動作正確度偵測。 資料前處理後,HomeXDL 從 CoP 軌跡以及壓力變化計算出的 特徵值,運用決策樹進行動 作種類的分辨,在深蹲、硬舉及弓箭步的動作分類中,準確度 (Accuracy)達91%以上。動作正確度判斷上,HomeXDL 運用 CNN 對上述三個動作的正確度偵測。在系統有標記的動作問題上, HomeXDL 也皆能快速且精準的偵測出來,三個動作整體準確度 也達90%以上。最後 HomeXDL 的使用者介面能及時的回饋給使 用者其運動品質與動作錯誤原因,對使用者在運動過程中有很大 的幫助。

    The pandemic of the coronavirus in 2019 has strong transmission power. Therefore, people are forced to stay at home and reduce going out. Sports centers, gyms and other places are also prone to increase the risk of infection due to sweating or close contact. Coupled with the recent popularity of sports, in order to maintain a good figure, people lose weight through exercise. However, most users tend to have bad movements or lack of systematic training without the assistance of a coach, resulting in sports injuries and limited effects. Therefore, this study proposes a Home Exercise Analysis System Based on Pressure Sensor and Deep Learning, called HomeXDL for short. HomeXDL uses pressure and pressure center of gravity changes combined with deep learning to achieve motion recording and motion accuracy detection. After data preprocessing, HomeXDL uses a decision tree to distinguish action types based on the CoP trajectory and feature values calculated from pressure changes. In the action classification of squats, deadlifts, and lunges, the accuracy rate exceeds 91%. At the same time, HomeXDL uses CNN to detect the correctness of the above three actions with over 90% overall accuracy rate. Finally, the user interface of HomeXDL can give real-time-feedback to the user on the quality of the movement and the cause of the error, which is of great help to the user during exercise.

    第一章 緒論 1 第二章 研究方法 3 第一節 壓力感應器相關研究 3 第二節 健身運動相關文獻探討 5 第三節 濾波器相關探討 8 第四節 卷積類神經網路背景 10 第三章 研究方法 16 第一節 資料的收集 17 第二節 資料前處理 19 第三節 運動姿勢分辨 24 第四節 動作分析紀錄與正確度偵測 28 第五節 使用者介面 38 第四章 實驗結果與分析 40 第一節 實驗環境 40 第二節 運動姿勢分辨結果探討 42 第三節 運動姿勢分辨結果探討 44 第四節 動作正確度分辨結果探討 48 第五章 結論與未來展望 62 參考文獻 63

    [1]  Kercher, Vanessa Marie, Kyle Kercher, Trevor Bennion, Brandon A. Yates, Yuri Feito, Chris Alexander, Paulo Costa Amaral, et al. “Fitness Trends from around the Globe.” ACSM’s Health and Fitness Journal 25, no. 1 (2021): 20–31. https://doi.org/10.1249/FIT.0000000000000639.
    [2]  L. Yang, Y. Li, D. Zeng and D. Wang, "Human Exercise Posture Analysis based on Pose Estimation," 2021 IEEE 5th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), 2021, pp. 1715-1719, doi: 10.1109/IAEAC50856.2021.9390870.
    [3]  Comfort, Paul, and Peter Kasim. “Optimizing Squat Technique.” Strength & Conditioning Journal 29, no. 6 (2007). https://journals.lww.com/nsca- scj/Fulltext/2007/12000/Optimizing_Squat_Technique.1.aspx
    [4]  懶人在家徒手健身大作戰,每天十分鐘,還你好身材。Accessed: March 8 , 2022 [online]. Available:https://www.worldgymtaiwan.com/training-blog-in/10-mins- freehand-home-workout
    [5]  不出門在家也能強化全身肌群的21個超實用訓練動作。Accessed: March 8 , 2022[online].Available:https://www.sportsplanetmag.com/article/desc/220006703
    [6]  12 種最適合減重的居家訓練動作。Accessed: March 8 2022 [online], Available: https://www.myprotein.tw/blog/training/12-exercises-for-weight-loss-at-home/
    [7]  Soter, Gabor, Martin Garrad, Andrew T. Conn, Helmut Hauser, and Jonathan Rossiter. “Skinflow: A Soft Robotic Skin Based on Fluidic Transmission.” In 2019 2nd IEEE International Conference on Soft Robotics (RoboSoft), 355–60, 2019. https://doi.org/10.1109/ROBOSOFT.2019.8722744.
    [8]  Kumar, Vaegae Naveen, and Komanapalli Venkata Lakshmi Narayana. “Development of an ANN-Based Pressure Transducer.” IEEE Sensors Journal 16, no. 1 (2016): 53–60. https://doi.org/10.1109/JSEN.2015.2477458.
    [9]  Strength Training,P.64-68, DK Publishing
    [10]  Peak Physique,P.78-80, Hollis Lance Liebman
    [11] 如何做出完美的弓箭步 運動星球, Accessed: June 10 2022 [online]. Available:https://www.sportsplanetmag.com/article/desc/16120716471062554, 2022/06/01
    [12]  Peak Physique,P.76, Hollis Lance Liebman
    [13]  Peak Physique,P.77, Hollis Lance Liebman
    [14]  Strength Training,P.85-88, DK Publishing
    [15]  Original: Kinesiology: Scientific Basis of Human Motion, Nancy Hamilton, Wendi Weimar, Kathryn Luttgens
    [16]  Chiu, Min-Chi, Hsin-Chieh Wu, and Li-Yu Chang. “Gait Speed and Gender Effects on Center of Pressure Progression during Normal Walking.” Gait & Posture 37, no. 1 (January 2013): 43–48. https://doi.org/10.1016/j.gaitpost.2012.05.030.
    [17]  Nagamune, Kouki, and Makoto Yamada. “A Wearable Measurement System for Sole Pressure to Calculate Center of Pressure in Sports Activity.” In 2018 IEEE International Conference on Systems, Man, and Cybernetics (SMC), 1333–36. IEEE Press, 2018. https://doi.org/10.1109/SMC.2018.00233.
    [18]  空間濾波 ,陳慶瀚
    [19]  Deng, G., and L.W. Cahill. “An Adaptive Gaussian Filter for Noise Reduction and Edge Detection.” In 1993 IEEE Conference Record Nuclear Science Symposium and Medical Imaging Conference, 1615–19 vol.3, 1993. https://doi.org/10.1109/ NSSMIC.1993.373563.
    [20]  Gallagher, Neal B. “Savitzky-Golay Smoothing and Differentiation Filter.” Eigenvector Research Incorporated, 2020.
    [21]  El-Sawy, Ahmed, Hazem M. El-Bakry, and Mohamed Loey. “CNN for Handwritten Arabic Digits Recognition Based on LeNet-5.” In AISI, 2016.
    [22]  Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E Hinton. “Imagenet Classification with Deep Convolutional Neural Networks.” Communications of the ACM 60, no. 6 (2017): 84–90.
    [23]  Wei, Yunchao, Wei Xia, Min Lin, Junshi Huang, Bingbing Ni, Jian Dong, Yao Zhao, and Shuicheng Yan. “HCP: A Flexible CNN Framework for Multi-Label Image Classification.” IEEE Transactions on Pattern Analysis and Machine Intelligence 38, no. 9 (September 2016): 1901–7. https://doi.org/10.1109/tpami.2015.2491929.
    [24]  Pascanu, Razvan, Tomas Mikolov, and Yoshua Bengio. “On the Difficulty of Training Recurrent Neural Networks.” arXiv, 2012. https://doi.org/10.48550/ARXIV.1211.5063.
    [25]  Quinlan, J. R. “Induction of Decision Trees.” Machine Learning 1, no. 1 (March 1, 1986): 81–106. https://doi.org/10.1007/BF00116251.

    無法下載圖示 本全文未授權公開
    QR CODE