ACXNet hybrid deep learning model for cross task mental workload estimation using EEG neural manifolds.

Area of Science:

• Neuroscience
• Cognitive Science
• Human-Computer Interaction

Background:

• Mental workload significantly impacts human performance and task efficiency.
• Traditional workload assessment methods (subjective reports, performance metrics) have limitations.
• Electroencephalography (EEG) offers an objective, continuous measure of cognitive states.

Purpose of the Study:

• To propose a novel hybrid approach, ACXNet, for accurate mental workload estimation using EEG.
• To develop a method that learns EEG features across tasks without subject-specific calibration or pre-labeled data.
• To enhance the robustness and precision of mental workload assessment for real-world applications.

Main Methods:

• ACXNet integrates an autoencoder for unsupervised feature extraction, a Convolutional Neural Network (CNN) for spatial-temporal dependencies, and XGBoost for classification.
• Utilized the STEW dataset with EEG recordings from 48 participants under varying cognitive loads.
• Employed a binary classification strategy to distinguish between low and high mental workload conditions.

Main Results:

• ACXNet achieved high accuracy: 92.10% for the SIMKAP task and 89.94% for the No task condition.
• The model demonstrated superior performance compared to existing methods in mental workload estimation.
• Successfully learned EEG features across tasks without requiring prior subject-specific calibration.

Conclusions:

• ACXNet offers a significant advancement in mental workload estimation, improving accuracy and robustness.
• The proposed method provides a scalable solution adaptable to diverse real-world scenarios.
• This research opens new possibilities for intelligent systems in human-computer interaction and healthcare.