Exploring EEG characteristics of multi-level mental stress based on human-machine system

Area of Science:

• Human-computer interaction
• Cognitive neuroscience
• Machine learning

Background:

• Assessing operator mental stress is crucial for developing effective human-machine systems (HMSs).
• Real-time monitoring of cognitive states during tasks presents significant challenges.
• Understanding neural correlates of mental stress is key to objective assessment.

Purpose of the Study:

• To investigate operator mental stress during a virtual unmanned vehicle (UAV) driving task.
• To develop a dataset of neural activity associated with varying mental stress levels.
• To create an automated system for mental stress assessment using electroencephalography (EEG) data.

Main Methods:

• A virtual UAV driving task with adjustable difficulty levels was designed.
• Real-time EEG data was collected from operators during the task.
• A multiple attention-based convolutional neural network (MACN) was developed for stress classification.

Main Results:

• EEG analysis revealed distinct neural patterns correlating with task difficulty and mental stress.
• Frontal theta power spectral density (PSD) decreased, while central beta-PSD increased with task difficulty.
• The MACN model achieved high classification accuracies (89.49% for arousal, 89.88% for valence).

Conclusions:

• Objective assessment of mental stress in HMSs is feasible using virtual task scenarios.
• The proposed MACN model offers a promising approach for real-time stress detection.
• Findings support advancements in cognitive computing and human-machine interaction applications.