Smartphone-based human fatigue level detection using machine learning approaches.

Area of Science:

• Biomechanics
• Wearable Technology
• Machine Learning

Background:

• Human muscle fatigue diminishes capability, increasing fall and injury risks.
• Objective fatigue monitoring is crucial for workplace safety and performance optimization.
• Current methods for fatigue assessment can be subjective or require specialized equipment.

Purpose of the Study:

• To develop and validate a machine learning model for classifying human fatigue levels.
• To utilize motion signals captured by a smartphone for fatigue detection.
• To assess the model's accuracy across different fatigue classification granularities.

Main Methods:

• Participants performed fatiguing exercises (squatting) followed by walking.
• A smartphone attached to the shank recorded gait data.
• Data was labeled using the Borg's Rating of Perceived Exertion (RPE).
• Machine learning models were trained to classify fatigue into two, three, and four levels.

Main Results:

• The two-level fatigue classification (no- vs. strong-fatigue) achieved 91% accuracy.
• Three-level (no-, medium-, strong-fatigue) and four-level (no-, low-, medium-, strong-fatigue) models reached 78% and 64% accuracy, respectively.
• The findings demonstrate the feasibility of using smartphone motion data for fatigue assessment.

Conclusions:

• A smartphone-based machine learning approach can effectively identify human muscle fatigue.
• This technology has the potential to be an accessible fatigue-monitoring tool in various settings.
• Implementing such tools may enhance worker performance and reduce the incidence of falls and injuries.