Bio-signals Collecting System for Fatigue Level Classification.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Human Factors Engineering

Background:

• Fatigue is a significant risk factor impacting quality of life, work efficiency, and safety in high-risk environments.
• Current fatigue assessment relies on subjective evaluation, lacking precise definition and quantification.
• Objective fatigue measurement is essential for risk management and accident prevention.

Purpose of the Study:

• To develop a machine learning and deep learning-based fatigue level classifier.
• To create a system for collecting and purifying bio-signal data for accurate fatigue level extraction.
• To establish a protocol for obtaining true fatigue levels, minimizing subjective bias.

Main Methods:

• Developed a bio-signal collection device to simultaneously capture visual, thermal, and vocal signals.
• Established a protocol for data acquisition and purification for accurate fatigue level extraction.
• Utilized the Daily Multidimensional Fatigue Inventory and physiological indicators to determine true fatigue levels, screening out subjective factors.

Main Results:

• Successfully gathered multi-modal bio-signal data (visual, thermal, vocal) for one-minute intervals.
• Established a protocol for purifying data to extract objective fatigue levels.
• Created a dataset of bio-signals linked with validated true fatigue levels for machine learning training.

Conclusions:

• Proposes a novel research methodology for objective fatigue assessment using bio-signals.
• Enables the training of machine learning and deep learning models for multi-level fatigue classification.
• Aims to improve safety, mission efficiency, and quality of life by providing a quantifiable fatigue evaluation tool.