Measurement and identification of mental workload during simulated computer tasks with multimodal methods and machine...

Area of Science:

• Human-Computer Interaction
• Physiological Computing
• Occupational Health

Background:

• Accurate mental workload measurement is crucial for optimizing performance and preventing burnout.
• Existing methods often rely on subjective or performance-based assessments, which can be limited.
• Developing objective, multimodal indicators for mental workload is an active area of research.

Purpose of the Study:

• To multimodally measure mental workload and validate physiological indicators for its estimation.
• To assess the efficacy of electrodermal activity (EDA) and respiration signals in detecting workload variations.
• To evaluate the accuracy of classification models using physiological data and task performance for workload detection.

Main Methods:

• A simulated computer task with varying difficulty levels was employed.
• Physiological signals (ECG, HRV, EMG, EDA, respiration), subjective ratings (NASA-TLX), and task performance were recorded.
• Statistical analyses and machine learning models (LDA, SVM, KNN, ANNs) were used for data analysis.

Main Results:

• Electrodermal activity (EDA) and respiration indices significantly increased with task difficulty.
• Average heart rate (AVHR) and LF/HF ratio did not show significant differences across tasks.
• Classification models achieved 96.4% accuracy using combined physiological and performance data, and 78.3% using physiological data alone.
• ECG and EDA signals demonstrated strong discriminating power for mental workload detection.

Conclusions:

• Multimodal measurement integrating physiological signals and task performance offers a highly accurate method for estimating mental workload.
• EDA and respiration are promising physiological indicators for real-time mental workload monitoring.
• Findings support the development of early detection systems for mental workload to improve worker health and efficiency.