Instantaneous mental workload assessment using time-frequency analysis and semi-supervised learning.

Area of Science:

• Human-Computer Interaction
• Machine Learning
• Cognitive Engineering

Background:

• Real-time mental workload (MWL) assessment is crucial for safety-critical intelligent human-machine cooperative systems.
• Data-driven machine learning (ML) approaches for MWL recognition face challenges due to insufficient labeled training data.

Purpose of the Study:

• To propose a novel semi-supervised extreme learning machine (SS-ELM) algorithm for MWL pattern classification.
• To address the limitation of requiring extensive labeled data in traditional ML models for MWL assessment.

Main Methods:

• Development and application of a semi-supervised extreme learning machine (SS-ELM) algorithm.
• Utilizing a small number of labeled data alongside a large volume of unlabeled data for model training.
• Classification of mental workload patterns.

Main Results:

• The proposed SS-ELM algorithm demonstrated significant improvements in the accuracy of MWL classification.
• The SS-ELM approach enhanced the efficiency of MWL recognition.
• Effective utilization of large unlabeled datasets was achieved.

Conclusions:

• The SS-ELM provides a competitive and effective machine learning approach for real-time MWL assessment.
• This method overcomes the data scarcity issue in training ML models for MWL recognition.
• The SS-ELM paradigm is suitable for real-world applications with abundant unlabeled data.