近年來，透過數據量化分析可以針對個別運動員的訓練成效給予評估，因此有研究提出使用影像記錄方式來擷取運動員訓練時的過程，但是影像記錄方式容易因攝影機拍攝角度範圍受限而受影響，進而影響球種辨識結果。
現今使用基於慣性感測元件的穿戴式裝置來監測運動已廣泛被使用，其優點有重量輕、功率低且易於攜帶，此方式能有效避免攝影機拍攝角度範圍受限問題，因此本論文開發一套基於慣性感測器的即時羽球擊球資訊系統，其系統可於電腦上即時顯示擊球數量、種類、速度及力道等資訊。此系統使用滑動視窗法結合本論文提出之演算法能準確的偵測擊球動作並透過SVM機器學習方式進行六種球種分類(挑球、撲球、平球、長球、切球和殺球)。我們希望比較慣性感測器在不同位置對於辨識影響差異，因此將無線慣性感測器分別裝設於羽球拍拍面、羽球拍拍底和受試者的右手手腕處來收集羽球選手在擊球動作時所產生的慣性數據，共收錄了10位受試者擊球動作數據。在離線分析方面，個人模型在三個位置(拍面、拍底、右手手腕)的平均準確度分別為97.24%、98.21%、97.94%，而通用模型在三個位置(拍面、拍底、右手手腕)的平均準確度分別為85.45%、89.32%、87.87%。在即時分析方面，個人模型在三個位置(拍面、拍底、右手手腕)的平均準確度分別為94.33%、94.67%、94.50%，而通用模型在三個位置(拍面、拍底、右手手腕)的平均準確度分別為85.27%、87.97%、84.97%。從上述結果可以發現，裝設於球拍拍底處之無線慣性感測器能得到相對較高的準確度。
此外，透過高速攝影機擷取擊球瞬間影像，結果發現裝設於羽球球拍拍面處之無線慣性感測器估算的羽球速度和高速攝影機估算的羽球速度有著相對較高的皮爾遜相關性r = 0.94。本論文所提出的系統能將擊球動作量化，讓使用者能即時知道自己的擊球動作資訊。

In recent years, quantitative data analysis can assess athletes' training performance. Therefore, some studies propose to use the image recording method to capture the training process of athletes. However, video recording can be affected by the limited angle of the camera, which in turn affects the results of stroke types identification.
Nowadays, wearable devices based on inertial sensor elements have been widely used to monitor sports. It has the advantages of being lightweight, low power, and easy portability. This method can effectively avoid the problem of limited camera angle range. Therefore, this thesis develops a real-time badminton stroke information system based on inertial sensors. The system can display the information on the stroke times, stroke types, stroke speed, and stroke force in real-time. This system uses the sliding window method combined with the proposed algorithm. It can detect the hitting action accurately and recognize the six stroke types (lift, rush, drive, clear, cut, and smash) through SVM machine learning. In offline analysis, the accuracy of the personal model was 97.24%, 98.21%, and 97.94% for the three positions (racket center, racket bottom, and right wrist), while the accuracy of the general model was 85.45%, 89.32%, and 87.87% for the three positions (racket center, racket bottom, and right wrist). In real-time analysis, the accuracy of the personal model was 94.33%, 94.67%, and 94.50% for the three positions (racket center, racket bottom, and right wrist), while the accuracy of the general model was 85.27%, 87.97%, and 84.97% for the three positions (racket center, racket bottom, and right wrist). The results show that the wireless inertial sensor installed at the racket bottom can obtain relatively high accuracy.
In addition, the high-speed camera was used to capture instant images of the stroke. The study found a high Pearson correlation r = 0.94 between the stroke speed estimated by the wireless inertial sensor installed in the racket center and the stroke speed estimated by the high-speed camera. The system proposed in this thesis can quantify the hitting action, allowing users to know their hitting action information in real-time.

[1] P. Oja et al., "Associations of Specific Types of Sports and Exercise with All-cause and Cardiovascular-disease Mortality: A Cohort Study of 80 306 British Adults," British Journal of Sports Medicine, vol. 51, no. 10, pp. 812-817, May 2017, doi: 10.1136/bjsports-2016-096822.
[2] K. Teu, W. Kim, J. Tan, and F. Fuss, "Using Dual Euler Angles for the Analysis of Arm Movement During the Badminton Smash," Sports Engineering, vol. 8, pp. 171-178, Jan. 2005, doi: 10.1007/BF02844017.
[3] A. Ghosh, S. Singh, and C. V. Jawahar, "Towards Structured Analysis of Broadcast Badminton Videos," 2018 IEEE Winter Conference on Applications of Computer Vision (WACV), pp. 296-304, Mar. 2018, doi: 10.1109/WACV.2018.00039.
[4] M. A. I. Anik, M. Hassan, H. Mahmud, and M. K. Hasan, "Activity Recognition of a Badminton Game Through Accelerometer and Gyroscope," 2016 19th International Conference on Computer and Information Technology (ICCIT), pp. 213-217, Dec. 2016, doi: 10.1109/ICCITECHN.2016.7860197.
[5] J. Lin et al., "Design and Implement a Mobile Badminton Stroke Classification System," 2017 19th Asia-Pacific Network Operations and Management Symposium (APNOMS), pp. 235-238, Sept. 2017, doi: 10.1109/APNOMS.2017.8094141.
[6] Y. Wang et al., "IoT for Next-Generation Racket Sports Training," IEEE Internet of Things Journal, vol. 5, no. 6, pp. 4558-4566, Dec. 2018, doi: 10.1109/JIOT.2018.2837347.
[7] H. Y. Ting, K. S. Sim, and F. Abas, "Kinect-based Badminton Movement Recognition and Analysis System," International Journal of Computer Science in Sport, vol. 14, pp. 25-41, Jan. 2015.
[8] K. Weeratunga, A. Dharmaratne, and K. B. How, "Application of Computer Vision and Vector Space Model for Tactical Movement Classification in Badminton," 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 132-138, Jul. 2017.
[9] N. P. Ju, D. R. Yu, T. U. İk, and W. C. Peng, "Trajectory-based Badminton Shots Detection," 2020 International Conference on Pervasive Artificial Intelligence (ICPAI), pp. 64-71, Dec. 2020, doi: 10.1109/ICPAI51961.2020.00020.
[10] Maradin, "Types and Advantages of MEMS Technology," https://www.maradin.co.il/advantages-of-mems-technology/(accessed Jan. 2022).
[11] Z. Wang, M. Guo, and C. Zhao, "Badminton Stroke Recognition Based on Body Sensor Networks," IEEE Transactions on Human-Machine Systems, vol. 46, no. 5, pp. 769-775, Oct. 2016, doi: 10.1109/THMS.2016.2571265.
[12] K. Xia et al., "Racquet Sports Recognition Using a Hybrid Clustering Model Learned from Integrated Wearable Sensor," Sensors, vol. 20, no. 6, pp. 1638, Mar. 2020, doi: 10.3390/s20061638.
[13] N. Ahmad et al., "Reviews on Various Inertial Measurement Unit (IMU) Sensor Applications," International Journal of Signal Processing Systems, vol. 1, no. 2, pp. 256-262, Dec. 2013, doi: 10.12720/ijsps.1.2.256-262.
[14] C. Pao, "Exploring the Application of Gyroscopes," https://www.ceva-dsp.com/ourblog/exploring-the-application-of-gyroscopes/(accessed Jan. 2022).
[15] W. R. Hamilton, "On Quaternions; Or on a New System of Imaginaries in Algebra," The London, Edinburgh, and Dublin Philosophical Magazine and 55 Journal of Science, vol. 25, no. 163, pp. 10-13, Jul. 1844, doi: 10.1080/14786444408644923.
[16] C. Cortes and V. Vapnik, "Support-vector Networks," Machine Learning, vol. 20, no. 3, pp. 273-297, Sept. 1995, doi: 10.1007/BF00994018.
[17] VICTOR勝利體育, "羽球教室," https://www.victorsport.com.tw/badmintonaz.
html?cid=139&order=latest(accessed Jan. 2022).
[18] Bluetooth, "Bluetooth Wireless Technology," https://www.bluetooth.com/learn-about-bluetooth/tech-overview/(accessed Jan. 2022).
[19] Nordic Semiconductor, "nRF52 DK," https://www.nordicsemi.com/Products/
Development-hardware/nrf52-dk(accessed Jan. 2022).
[20] Freepik, "Badminton Silluette Clipart Badminton Silhouette Sports," https://www.pinclipart.com/pindetail/iRhohh_badminton/silluetteclipart%20bad-minton-silhouette-sports-badminton-player/(accessed Jan. 2022).
[21] Nordic Semiconductor, "nRF52832," https://www.nordicsemi.com/Products/Low
-power-short-range-wireless/nRF52832(accessed Jan. 2022).
[22] STMicroelectronics, "H3LIS331DL," https://www.st.com/resource/en/datasheet/
h3lis331dl.pdf(accessed Jan. 2022).
[23] InvenSense, "ICM-20649," https://product.tdk.com/system/files/dam/doc/product
/sensor/mortion-inertial/imu/data_sheet/ds-000192-icm-20649-v1.0.pdf(accessed Jan. 2022).
[24] F. Pedregosa et al., "Scikit-learn: Machine Learning in Python," J. Mach. Learn. Res, vol. 12, pp. 2825-2830, Oct. 2011.
[25] Scikit-learn, "Scikit-learn Algorithm Cheat-sheet," https://scikit-learn.org/stable/
tutorial/machine_learning_map/index.html(accessed Jan. 2022).
[26] A. G. Asuero, A. Sayago, and A. G. González, "The Correlation Coefficient: An Overview," Critical Reviews in Analytical Chemistry, vol. 36, no. 5, pp. 41-59, Jan. 2007, doi: 10.1080/10408340500526766.
[27] 蔡怡娟, "基於手機APP的互動式羽球訓練輔助系統," 國立台灣科技大學碩士論文, 2021.
[28] 楊佳玲, "基於球拍感測器與機器學習的羽球擊球資訊系統," 國立台灣科技大學碩士論文, 2021.
[29] 黃正瑜, "基於智慧手錶的即時羽球球種辨識," 國立台灣科技大學碩士論文, 2021.