@inproceedings{oai:ipsj.ixsq.nii.ac.jp:00213082,
 author = {花田, 祥典 and 横窪, 安奈 and ロペズ, ギヨーム},
 book = {マルチメディア，分散協調とモバイルシンポジウム2021論文集},
 issue = {1},
 month = {Jun},
 note = {Maintaining healthy living requires habitual physical activities.  Nonetheless, staying motivated to work out regularly is challenging for most people.  To solve this problem, automated personal supporting systems could help.  This paper presents boxercise, a fitness standard exercise that mainly includes shadow-boxing exercises.  The paper introduces punch activity detection and classification methods using acceleration and angular velocity signals recorded using a single smartwatch on the participant’s rear hand wrist.  The proposed method is evaluated on our 10 participants aged between 17 and 53 years old (8 male and 2 female, age  27.8±12.8).   As  a  result,  we  achieved  98.8%  detection  accuracy,  98.9%  classification  accuracy  with SVM  in-person-dependent  (PD)  case,  and  91.1%  classification  accuracy  with  SVM  in  person-independent (PI) case.  In addition, we estimated the real-time performance of each classification method and found out all our methods could classify a single punch in less than 0.1 seconds.  The paper also discussed some points of improvement towards a practical boxercise supporting system., Maintaining healthy living requires habitual physical activities.  Nonetheless, staying motivated to work out regularly is challenging for most people.  To solve this problem, automated personal supporting systems could help.  This paper presents boxercise, a fitness standard exercise that mainly includes shadow-boxing exercises.  The paper introduces punch activity detection and classification methods using acceleration and angular velocity signals recorded using a single smartwatch on the participant’s rear hand wrist.  The proposed method is evaluated on our 10 participants aged between 17 and 53 years old (8 male and 2 female, age  27.8±12.8).   As  a  result,  we  achieved  98.8%  detection  accuracy,  98.9%  classification  accuracy  with SVM  in-person-dependent  (PD)  case,  and  91.1%  classification  accuracy  with  SVM  in  person-independent (PI) case.  In addition, we estimated the real-time performance of each classification method and found out all our methods could classify a single punch in less than 0.1 seconds.  The paper also discussed some points of improvement towards a practical boxercise supporting system.},
 pages = {1304--1309},
 publisher = {情報処理学会},
 title = {IMUセンサーを用いたパンチ検出と分類手法の提案},
 volume = {2021},
 year = {2021}
}