政大機構典藏-National Chengchi University Institutional Repository(NCCUR):Item 140.119/155455
English  |  正體中文  |  简体中文  |  Post-Print筆數 : 27 |  全文筆數/總筆數 : 114396/145431 (79%)
造訪人次 : 53048157      線上人數 : 391
RC Version 6.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
搜尋範圍 查詢小技巧：
  • 您可在西文檢索詞彙前後加上"雙引號"，以獲取較精準的檢索結果
  • 若欲以作者姓名搜尋，建議至進階搜尋限定作者欄位，可獲得較完整資料
    •  進階搜尋
    政大機構典藏 > 資訊學院 > 資訊科學系 > 學位論文 >  Item 140.119/155455


    請使用永久網址來引用或連結此文件: https://nccur.lib.nccu.edu.tw/handle/140.119/155455


    題名: FedTS：基於時間序列聯邦學習的羽球揮拍評分
    FedTS: Federated Learning Based on Time Series For Scoring Badminton Strokes
    作者: 思沛淇
    Si, Pei-Qi
    貢獻者: 蔡子傑
    Tsai, Tzu-Chieh
    思沛淇
    Si, Pei-Qi
    關鍵詞: 聯邦學習
    非獨立同分佈
    時間序列資料
    穿戴式裝置
    知識蒸餾
    Federated Learning
    Non-iid
    Time-series Data
    Wearable Devices
    Knowledge Distillation
    日期: 2025
    上傳時間: 2025-02-04 15:44:28 (UTC+8)
    摘要: 近年來，隨著疫情逐漸趨緩，人們開始積極地投入各項戶外運動。其中，羽球運動因戴資穎在國際賽的亮眼表現，以及李洋、王齊麟勇奪奧運金牌，一時間成為臺灣運動愛好者的矚目焦點，帶動羽球運動人口持續攀升。根據教育部體育署相關調查，網路族群之中每三人便有一人使用穿戴式裝置紀錄運動數據；穿戴式裝置在運動科技領域已成為一種時尚趨勢。

    本研究針對時間序列資料的特性，運用穿戴式裝置（Fitbit 智慧手環）收集羽球揮拍資訊，包括三軸加速度與三軸角速度，透過程式串接 Fitbit 提供的 API 即可輕鬆取得。與多數應用影像技術或在球拍嵌入晶片的既有研究相比，穿戴式裝置更具有隨手可得、操作便利的優勢。然而，目前尚缺乏以時間序列為基礎的聯邦學習（Federated Learning, FL）應用於羽球運動評分之研究，故本研究特別關注在「非獨立同分佈（non-iid）資料下的聯邦學習訓練」，以提升模型對實際使用者情境的適應性。

    本研究旨在：（1）實際收集揮拍時間序列數據並標註五個面向的教練評分；（2）探討在此資料集中可能面臨的 non-iid 問題，並比較傳統聯邦學習（FedAvg）以及其他改良演算法（FedProx、FedBN、FedTS）在不同實驗情境下之表現；（3）提供各種增量式資料訓練分析，檢驗在初期資料有限、或客戶端分佈差異極大的真實情境中，如何利用蒸餾技術（Knowledge Distillation）與動態正則項（Proximal Term）等方式改善模型效能。最終，研究成果將去識別化後公開於網路上，期能成為未來進行運動數據科學研究之基礎。
    In recent years, as the COVID-19 pandemic has gradually subsided, people have increasingly engaged in various outdoor activities. Among these, badminton has gained significant popularity in Taiwan, driven by Tai Tzu-ying's impressive performances in international competitions and the Olympic gold medal won by Lee Yang and Wang Chi-lin. The growing interest in badminton has led to a rising number of enthusiasts. According to surveys conducted by the Sports Administration of the Ministry of Education, one in three internet users employs wearable devices to track their fitness data, making wearable technology a prominent trend in sports science.

    This study leverages the characteristics of time-series data by collecting badminton stroke information using wearable devices (Fitbit smart bands), including tri-axial acceleration and tri-axial angular velocity. Through API integration provided by Fitbit, data can be effortlessly obtained. Compared to existing studies that primarily utilize imaging technology or embed sensors into rackets, wearable devices offer a more accessible and convenient solution. However, there is a lack of research applying Federated Learning (FL) based on time-series data for badminton performance scoring. This study focuses specifically on Federated Learning training under non-independent and identically distributed (non-iid) data to enhance model adaptability to real-world user scenarios.

    The objectives of this research are: (1) to collect real-time stroke time-series data and annotate them with coach evaluations across five dimensions; (2) to investigate the challenges of non-iid data in this dataset and compare the performance of traditional Federated Learning (FedAvg) with other advanced algorithms (FedProx, FedBN, FedTS) under different experimental scenarios; and (3) to analyze incremental data training, examining how techniques such as knowledge distillation and dynamic regularization (Proximal Term) can improve model performance in real-world scenarios with limited initial data or significant client distribution disparities. The research results, anonymized and de-identified, will be made publicly available online, aiming to serve as a foundation for future studies in sports data science.
    參考文獻: 1] Fabrizio de Fabritiis and Konstantinos Gryllias. A federated learning approach for rolling bearing fault diagnosis on data sources with imbalanced class distribution. In Surveillance, Vibrations, Shock and Noise, 2023.
    [2] 王威堯 (Wei-Yao Wang), 張凱翔 (Kai-Shiang Chang), 陳霆峰 (Teng-Fong Chen), 王志全 (Chih-Chuan Wang), 彭文志 (Wen-Chih Peng), and 易志偉 (Chih-Wei Yi). Badminton coach ai：基於深度學習之羽球賽事資訊分析平台. 體育學報, 53(2):201–213, Jun 2020.
    [3] Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Agueray Arcas. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pages 1273–1282. PMLR, 2017.
    [4] Peter Kairouz, H Brendan McMahan, Brendan Avent, Aurélien Bellet, Mehdi Bennis, Arjun Nitin Bhagoji, Kallista Bonawitz, Zachary Charles, Graham Cormode, Rachel Cummings, et al. Advances and open problems in federated learning. Foundations and trends® in machine learning, 14(1–2):1–210, 2021.
    [5] Qinbin Li, Yiqun Diao, Quan Chen, and Bingsheng He. Federated learning on non-iid data silos: An experimental study. In 2022 IEEE 38th international conference on data engineering (ICDE), pages 965–978. IEEE, 2022.
    [6] Tian Li, Anit Kumar Sahu, Manzil Zaheer, Maziar Sanjabi, Ameet Talwalkar, and Virginia Smith. Federated optimization in heterogeneous networks. Proceedings of Machine learning and systems, 2:429–450, 2020.
    [7] Xiaoxiao Li, Meirui Jiang, Xiaofei Zhang, Michael Kamp, and Qi Dou. Fedbn: Federated learning on non-iid features via local batch normalization. arXiv preprint arXiv:2102.07623, 2021.
    [8] Geoffrey Hinton. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531, 2015.
    [9] Tao Lin, Lingjing Kong, Sebastian U Stich, and Martin Jaggi. Ensemble distillation for robust model fusion in federated learning. Advances in neural information pro-
    cessing systems, 33:2351–2363, 2020.
    描述: 碩士
    國立政治大學
    資訊科學系
    111753214
    資料來源: http://thesis.lib.nccu.edu.tw/record/#G0111753214
    資料類型: thesis
    顯示於類別:[資訊科學系] 學位論文

    文件中的檔案:

    檔案 描述 大小格式瀏覽次數
    321401.pdf2551KbAdobe PDF0檢視/開啟


    在政大典藏中所有的資料項目都受到原著作權保護.


    社群 sharing

    著作權政策宣告 Copyright Announcement
    1.本網站之數位內容為國立政治大學所收錄之機構典藏，無償提供學術研究與公眾教育等公益性使用，惟仍請適度，合理使用本網站之內容，以尊重著作權人之權益。商業上之利用，則請先取得著作權人之授權。
    The digital content of this website is part of National Chengchi University Institutional Repository. It provides free access to academic research and public education for non-commercial use. Please utilize it in a proper and reasonable manner and respect the rights of copyright owners. For commercial use, please obtain authorization from the copyright owner in advance.

    2.本網站之製作，已盡力防止侵害著作權人之權益，如仍發現本網站之數位內容有侵害著作權人權益情事者，請權利人通知本網站維護人員(nccur@nccu.edu.tw)，維護人員將立即採取移除該數位著作等補救措施。
    NCCU Institutional Repository is made to protect the interests of copyright owners. If you believe that any material on the website infringes copyright, please contact our staff(nccur@nccu.edu.tw). We will remove the work from the repository and investigate your claim.
    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - 回饋