Exercise fatigue diagnosis method based on short-time Fourier transform and convolutional neural network.

Area of Science:

• Sports Science
• Biomedical Engineering
• Artificial Intelligence

Background:

• Determining safe exercise limits is challenging for athletes, risking physical damage due to overexertion.
• Preventing exercise-related injuries requires reliable methods to assess an individual's fatigue level.
• Current methods may not provide real-time, accurate fatigue status during physical activity.

Purpose of the Study:

• To develop and validate an exercise fatigue diagnosis method using electrocardiogram (ECG) signals.
• To enhance sports safety by enabling timely detection of non-optimal exercise states.
• To leverage advanced signal processing and machine learning for accurate fatigue assessment.

Main Methods:

• Applying short-time Fourier transform (STFT) to ECG signals to generate time-spectrum data.
• Utilizing a convolutional neural network (CNN) for machine learning on the ECG time-spectrum data.
• Training and validating the CNN model using ECG data from exercise fatigue experiments.

Main Results:

• The developed method achieved a high recognition accuracy rate of 97.70% in diagnosing exercise fatigue.
• Experimental validation confirmed the feasibility and effectiveness of the STFT-CNN approach.
• The model successfully identified exercise fatigue levels from real-time ECG signals.

Conclusions:

• The proposed STFT-CNN method offers a highly accurate and feasible approach for diagnosing exercise fatigue.
• This technology can significantly contribute to preventing sports injuries by monitoring exerciser fatigue.
• Real-time ECG analysis provides a promising avenue for personalized sports safety and performance optimization.