Related Experiment Video
Updated: Dec 11, 2025
06:49
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
Published on: December 11, 2015
9.2K
Badminton Activity Recognition Using Accelerometer Data
Tim Steels1, Ben Van Herbruggen1, Jaron Fontaine1
1IDLab, Department of Information Technology, Ghent University-imec, 9000 Ghent, Belgium.
Sensors (Basel, Switzerland)
|August 23, 2020
Summary
Analyzing badminton movements is crucial for player development. This study uses low-cost sensors and a novel neural network to classify nine badminton activities with high precision, offering an accessible alternative to video analysis.
Area of Science:
- Sports Science
- Biomechanical Analysis
- Machine Learning in Sports
Background:
- Traditional video analysis for badminton training is expensive and intrusive.
- Accurate analysis of player movements is vital for performance enhancement.
- There is a need for cost-effective and user-friendly motion analysis tools.
Purpose of the Study:
- To classify badminton movements using accelerometer and gyroscope data.
- To develop a novel neural network for activity recognition in badminton.
- To provide a low-cost, easy-to-use solution for badminton game analysis.
Main Methods:
- Collected data using off-the-shelf accelerometer and gyroscope sensors.
- Designed and implemented a novel convolutional neural network with variable frame sizes.
- Organized a data capturing campaign to record badminton movements.
- Evaluated the impact of sensor placement and sampling frequencies.
Main Results:
- The novel neural network achieved 86% precision classifying nine activities using only accelerometer data at 50 Hz.
- Incorporating gyroscope data increased precision to 99%.
- The proposed method outperformed traditional convolutional neural networks (79% and 88% precision).
Conclusions:
- The developed system offers a low-cost and user-friendly method for analyzing badminton games.
- Sensor-based motion classification provides valuable insights for training and performance.
- The novel neural network architecture effectively distinguishes various badminton activities.
