Scholarlink

This study presents a machine learning-based framework for the classification and repetition counting of barbell exercises using wearable sensor data. MetaMotion devices equipped with accelerometers and gyroscopes were used to capture movement signals across multiple strength training sessions. The dataset underwent systematic preprocessing, feature extraction, and predictive modeling to ensure reliable analysis. Several machine learning algorithms—including Random Forest, Support Vector Machines, Decision Trees, and k-Nearest Neighbors—were evaluated for their ability to accurately classify exercise types and count repetitions. The proposed framework achieved classification accuracies exceeding 90%, with certain models reaching up to 98%. In addition, the repetition-counting module demonstrated error margins below 2%. These results highlight the potential of sensor-driven machine learning systems for precise workout monitoring, offering practical applications in personalized fitness tracking, rehabilitation programs, and performance optimization. This research contributes to the field of human activity recognition by establishing a reliable and scalable approach for automated exercise detection and quantification.

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

Machine Learning-Based Classification and Repetition Counting of Barbell Exercises Using Wearable Sensor Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Bhoomika Gupta,
  • Mannya Agrawal,
  • Aaryan Gupta,
  • Preeti Narooka,
  • Parul Madan

摘要

This study presents a machine learning-based framework for the classification and repetition counting of barbell exercises using wearable sensor data. MetaMotion devices equipped with accelerometers and gyroscopes were used to capture movement signals across multiple strength training sessions. The dataset underwent systematic preprocessing, feature extraction, and predictive modeling to ensure reliable analysis. Several machine learning algorithms—including Random Forest, Support Vector Machines, Decision Trees, and k-Nearest Neighbors—were evaluated for their ability to accurately classify exercise types and count repetitions. The proposed framework achieved classification accuracies exceeding 90%, with certain models reaching up to 98%. In addition, the repetition-counting module demonstrated error margins below 2%. These results highlight the potential of sensor-driven machine learning systems for precise workout monitoring, offering practical applications in personalized fitness tracking, rehabilitation programs, and performance optimization. This research contributes to the field of human activity recognition by establishing a reliable and scalable approach for automated exercise detection and quantification.