羽球已成為熱門運動項目之一，為了有效的記錄和分析羽球球路，大多採用動作感測裝置及攝影機錄影。由於錄影的方式受限於拍攝角度與範圍的影響，較無法即時分辨出各種羽球的動作。然而使用動作感測裝置量測，除了能直觀的分析揮擊變化，進而改善影像處理的缺陷，且慣性感測器同時具有低成本及體積小的優點。
因此，本論文使用加速度計、陀螺儀及磁力計作為動作感測裝置，將裝置安裝於球拍拍桿上，並錄製球員揮擊時的資料。本論文收錄的動作包括正反手之長球、切球、平球、挑球及撲球等共10種動作。因球拍的兩面皆能揮擊，可能導致初始定義的軸向不同而造成誤判，因此本論文使用四元數及座標轉換將裝置軸向映射至地磁方向，以解決轉拍造成的影響。
於分類中本論文使用Sequential minimal optimization (Polynomial Kernel)作為動作分類器，能夠辨識10種揮擊的動作。於離線分析中，本論文使用之辨識方式準確度能高達98.75%。於即時辨識中，準確度為95.17%，高於市售產品，表示本論文開發之系統具有相當的優勢。
本論文開發之系統除了解決轉拍的影響，可以準確的辨識正反手共10種球種，也能夠即時的分析及記錄個人揮擊動作，相較於傳統利用錄影的方式，可達到輕便及即時的效果。

Badminton has become one of the popular sports, in order to effectively record and analyze the activities of badminton strokes, most of the use of motion sensing devices and video camera. Due to the way the video is limited by the angle and range of the camera, it is not easy to classify the types of badminton strokes in real time. However, the measurement of motion sensing device can improve the defect of image processing in addition to the intuitive analyze of badminton stroke change, and the inertial sensor has the advantages of low cost and small volume.
Therefore, in this paper, the accelerometer, gyroscope and magnetometer are used as motion sensing devices to install the device on the racket and record the data when the player strokes. The activities of badminton strokes recorded in this paper includes 10 kinds of activities, such as clear, drop, drive, lift and netshot of forehand and backhand. But the two faces of the racket both can stroke, which may cause the initial axial directions are different, this paper combine quaternions with coordinate transformations, and make the acceleration of the racket to the geomagnetic acceleration, and reduce the effects from rotating racket.
In the classification, this paper uses sequential minimal optimization (polynomial Kernel) as activities classifier, can classify 10 kinds of activities, in the off-line analyze, this paper uses the classification method accuracy can be as high as 98.75%. In real time recognition, the accuracy is 95.17%, higher than the commercial products, indicating that the system developed in this paper has a considerable advantage.
In addition to solving the effects of the rotating racket, the system can classify 10 kinds of badminton strokes accurately, and can analyze and record the individual strokes in real time, which can achieve the light and real-time effect compared with the traditional video recording method.

[1] Gallup Market Research Corp, “106年運動現況調查結案報告,” 2017.
[2] 教育部體育署, “民眾申請加入特定體育團體會員報名專區.” [Online]. Available: https://www.sa.gov.tw/wSite/mp?mp=11.
[3] 友练USENSE, “羽毛球传感器.” [Online]. Available: http://www.ubc-tech.com/badminton.html.
[4] KOOSPUR酷浪, “酷浪小羽3.0.” [Online]. Available: http://www.coollang.com/welcome/badminton3.
[5] M. S.Salim, H. N.Lim, M. S. M.Salim, andM. Y.Baharuddin, “Motion analysis of arm movement during badminton smash,” in Proceedings of 2010 IEEE EMBS Conference on Biomedical Engineering and Sciences, IECBES 2010, 2010, no. December, pp. 111–114.
[6] S.Ramasinghe, K. G. M.Chathuramali, andR.Rodrigo, “Recognition of badminton strokes using dense trajectories,” in 2014 7th International Conference on Information and Automation for Sustainability: “Sharpening the Future with Sustainable Technology”, ICIAfS 2014, 2014.
[7] K.Weeratunga, A.Dharmaratne, andK. B.How, “Application of Computer Vision and Vector Space Model for Tactical Movement Classification in Badminton,” in IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, 2017, vol. 2017–July, pp. 132–138.
[8] M. A. I.Anik, M.Hassan, H.Mahmud, andM. K.Hasan, “Activity Recognition of a Badminton Game Through Accelerometer and Gyroscope,” in 19th International Conference on Computer and Information Technology, ICCIT 2016, 2016, pp. 213–217.
[9] Z.Wang, C.Zhao, andM.Guo, “Badminton Stroke Recognition Based on Body Sensor Networks,” Ieee Trans. Human-Machine Syst., vol. 46, no. 5, pp. 769–775, 2016.
[10] 張家偉, “智慧羽球拍應用於球種分類,” 國立交通大學, 2016.
[11] 陳姮妤, “基於混合式DTW之羽球動作辨識演算法開發,” 國立臺灣科技大學, 2017.
[12] C. Z.Shan, E. S. L.Ming, H. A.Rahman, andY. C.Fai, “Investigation of Upper Limb Movement during Badminton Smash,” in 2015 10th Asian Control Conference: Emerging Control Techniques for a Sustainable World, ASCC 2015, 2015, pp. 1–6.
[13] M. I.Rusydi, M.Sasaki, M. H.Sucipto, Zaini, andN.Windasari, “Local Euler Angle Pattern Recognition for Smash and Backhand in Badminton Based on Arm Position,” Procedia Manuf., vol. 3, no. Ahfe, pp. 898–903, 2015.
[14] M. I.Rusydi, M.Sasaki, M. H.Sucipto, Zaini, andN.Windasari, “Study about backhand short serve in badminton based on the Euler angle,” Proc. - 2015 4th Int. Conf. Instrumentation, Commun. Inf. Technol. Biomed. Eng. ICICI-BME 2015, pp. 108–112, 2015.
[15] M. I.Rusydi, S.Huda, F.Rusydi, M. H.Sucipto, andM.Sasaki, “Pattern recognition of overhead forehand and backhand in badminton based on the sign of local Euler angle,” Indones. J. Electr. Eng. Comput. Sci., vol. 2, no. 3, pp. 625–635, 2016.
[16] W.Pei, J.Wang, X.Xu, Z.Wu, andX.Du, “An embedded 6-axis sensor based recognition for tennis stroke,” 2017 IEEE Int. Conf. Consum. Electron. ICCE 2017, pp. 55–58, 2017.
[17] W.-F.Wang, C.-Y.Yang, andD.-Y.Wang, “Analysis of Movement Effectiveness in Badminton Strokes with Accelerometers,” in Advances in Intelligent Systems and Computing, 2016, vol. 387, pp. 95–104.
[18] S. O. H.Madgwick, A. J. L.Harrison, andR.Vaidyanathan, “Estimation of IMU and MARG orientation using a gradient descent algorithm,” in 2011 IEEE International Conference on Rehabilitation Robotics, 2011, pp. 1–7.
[19] I. H.Witten, E.Frank, andM. A.Hall, Data Mining:Practical Machine Learning Tools and Techniques. 2011.
[20] M.Lemkin andB. E.Boser, “A Three-Axis Micromachined Accelerometer with a CMOS Position-Sense Interface and Digital Offset-Trim Electronics,” IEEE J. Solid-State Circuits, vol. 34, no. 4, pp. 456–468, 1999.
[21] Huikai Xie andGary K. Fedder, “Fabrication, Characterization, and Analysis of a DRIE CMOS-MEMS Gyroscope,” IEEE Sens. J., vol. 3, no. 5, pp. 622–631, 2003.
[22] K.Mohamadabadi, Anisotropic magnetoresistance Magnetometer for inertial navigation systems. 2014.
[23] R.Mahony, T.Hamel, andJ.-M.Pflimli, “Nonlinear Complementary Filters on the Special Orthogonal Group,” IEEE Trans. Automat. Contr., vol. 53, no. 5, pp. 1203–1218, 2008.
[24] D. R.Wilkins, “ON QUATERNIONS By William Rowan Hamilton,” Proc. R. Irish Acad., vol. 3, no. 1847, pp. 1–16, 1999.
[25] J. B.Kuipers, “Quaternions and Rotation Sequences,” Geom. Integr. Quantization, pp. 127–143, 2000.
[26] J. C.Platt, “Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines,” Adv. kernel methods, pp. 185–208, 1998.
[27] L.Breiman, “Random Forests,” pp. 1–33, 2001.
[28] A.McCallum andK.Nigam, “A Comparison of Event Models for Naive Bayes Text Classification,” AAAI/ICML-98 Work. Learn. Text Categ., pp. 41–48, 1998.
[29] B. E.Boser, I. M.Guyon, andV. N.Vapnik, “A Training Algorithm for Optimal Margin Classiiers,” Proc. fifth Annu. Work. Comput. Learn. theory, pp. 144–152, 1992.
[30] VICTOR 勝利體育, “羽球教室.” [Online]. Available: http://www.victorsport.com.tw/badmintonaz.html?cid=139&order=latest.
[31] Nordic Semiconductor, “nRF52832 Multi-protocol Bluetooth Low Energy and 2.4GHz proprietary system-on-chip,” no. Bluetooth 5, pp. 0–1.
[32] InvenSense, “MPU9250 Product Specification Revision 1.1,” Prod. Specif., vol. 1, no. 408, pp. 1–4, 2016.
[33] J.Diebel, Representing attitude: Euler angles, unit quaternions, and rotation vectors, vol. 58. 2006.