Pattern Recognition of Momentary Mental Workload Based on Multi-Channel Electrophysiological Data and Ensemble...

Area of Science:

• * Neuroscience and Cognitive Science
• * Machine Learning and Artificial Intelligence

Background:

• * Accurate classification of mental workload (MWL) is crucial for understanding cognitive states.
• * Traditional machine learning methods often require manual feature engineering.
• * Convolutional Neural Networks (CNNs) show promise but require optimization for MWL tasks.

Purpose of the Study:

• * To optimize Convolutional Neural Networks (CNNs) for mental workload classification using physiological data.
• * To develop an Ensemble Convolutional Neural Network (ECNN) for enhanced MWL classification accuracy and robustness.
• * To compare the performance of ECNN against individual CNN models and traditional methods.

Main Methods:

• * Investigated optimal depth and parameter optimization algorithms for base CNN models.
• * Developed an Ensemble Convolutional Neural Network (ECNN) using weighted averaging, majority voting, and stacking.
• * Employed a resampling strategy to increase the diversity of individual CNN models.

Main Results:

• * Evaluated base CNNs on Accuracy, Precision, F-measure, G-mean, and training time.
• * The proposed ECNN framework demonstrated superior MWL classification performance compared to individual CNNs.
• * ECNN achieved entirely automatic feature extraction and MWL classification.

Conclusions:

• * The ECNN framework effectively enhances mental workload classification performance.
• * ECNN offers a robust and automated approach for analyzing physiological data for MWL.
• * This method surpasses traditional machine learning techniques in automatic feature extraction and classification.