Fatigue Prediction in Outdoor Running Conditions using Audio Data.

Area of Science:

• Sports Medicine
• Biomedical Engineering
• Machine Learning

Background:

• Running is a popular activity, but overuse injuries affect 29-79% of runners annually.
• Fatigue is a primary cause of these injuries, altering running biomechanics.
• Current fatigue monitoring methods can be invasive or impractical for real-world use.

Purpose of the Study:

• To investigate the feasibility of using smartphone-acquired audio data to model perceived exertion in runners.
• To explore the potential of machine learning, specifically Convolutional Neural Networks (CNNs), for fatigue estimation.

Main Methods:

• Audio data was collected from smartphones attached to runners' arms in outdoor settings.
• Log-Mel spectrograms were generated from the audio recordings.
• CNNs were employed to analyze spectrograms and predict the Borg Rating of Perceived Exertion (RPE) scale.

Main Results:

• The study achieved a Mean Absolute Error (MAE) of 2.35 in subject-dependent fatigue modeling.
• This demonstrates that audio signals can effectively capture and model physiological fatigue.
• The approach offers a non-invasive and accessible alternative to traditional sensor-based methods.

Conclusions:

• Audio data, processed with CNNs, can reliably model runner fatigue.
• This non-invasive method provides a practical tool for monitoring exertion and potentially preventing running injuries.
• Smartphone-based audio analysis presents a promising avenue for accessible athlete monitoring.