Scholarlink

Badminton is a popular sport that requires players to possess a diverse set of skills and techniques. Traditional methods of stroke classification rely heavily on human observation and subjective judgment, leading to inconsistencies and limitations. In order to improve player performance and provide personalized training, the classification of different badminton strokes is crucial. We propose a novel approach using machine learning techniques based on IMU and pressure sensor data. A sensor system integrated into the racket handle captures precise information about movements and racket dynamics. Machine learning algorithms i.e. Support Vector Machine (SVM) and k-Nearest Neighbours (kNN) are used to predict stroke types using features extracted from sensor data and compared based on performance metrics The SVM model achieved an accuracy of 96.9%, while the kNN model achieved an accuracy of 95.8%. The marginal difference of 1.1% suggests that SVM exhibits a slight advantage over kNN in accurately classifying badminton strokes and assessing player performance, making it a slightly more promising choice for this specific application.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Classification of Badminton Strokes Through IMU Data Using Supervised SVM and kNN Machine Learning Algorithms

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wan Hasbullah Mohd Isa,
  • Muhamad Irham Najmi Mohd Nadzri,
  • Nur Aliya Syahirah Binti Badrol Hisam,
  • Muhammad Amirul Abdullah,
  • Mohd Azraai Mohd Razman,
  • Anwar P. P. Abdul Majeed,
  • Ismail Mohd Khairuddin

摘要

Badminton is a popular sport that requires players to possess a diverse set of skills and techniques. Traditional methods of stroke classification rely heavily on human observation and subjective judgment, leading to inconsistencies and limitations. In order to improve player performance and provide personalized training, the classification of different badminton strokes is crucial. We propose a novel approach using machine learning techniques based on IMU and pressure sensor data. A sensor system integrated into the racket handle captures precise information about movements and racket dynamics. Machine learning algorithms i.e. Support Vector Machine (SVM) and k-Nearest Neighbours (kNN) are used to predict stroke types using features extracted from sensor data and compared based on performance metrics The SVM model achieved an accuracy of 96.9%, while the kNN model achieved an accuracy of 95.8%. The marginal difference of 1.1% suggests that SVM exhibits a slight advantage over kNN in accurately classifying badminton strokes and assessing player performance, making it a slightly more promising choice for this specific application.