A Multi-Scale Adaptive EEG Feature Selection and Weighting Method for Mental Workload Estimation in Rapid Serial...

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Accurate mental workload estimation is vital for Brain-Computer Interface (BCI) performance, especially in Rapid Serial Visual Presentation (RSVP) systems.
• Existing Electroencephalography (EEG)-based methods face challenges with high-dimensional data, feature redundancy, and limited generalization.

Purpose of the Study:

• To develop an advanced EEG-based framework for accurate mental workload estimation in RSVP-BCI.
• To address limitations of existing methods by improving feature selection and representation.

Main Methods:

• Designed an RSVP-based aircraft target detection task with varying presentation rates to induce distinct workload levels.
• Collected behavioral, subjective, and high-density EEG data.
• Proposed a Multi-scale Adaptive Feature selection and Augmented Weighting (MAFA) framework incorporating adaptive channel selection and multi-scale feature compression/weighting.

Main Results:

• The multi-rate RSVP paradigm successfully elicited different workload levels with significant EEG feature variations.
• The MAFA framework achieved higher classification accuracy compared to existing methods.
• Feature weight visualization indicated adaptive attention to posterior EEG regions (theta and alpha bands), aligning with workload-related brain patterns.

Conclusions:

• The MAFA framework demonstrates feasibility and interpretability for mental workload estimation.
• The proposed method shows potential for enhancing accuracy and generalization in RSVP-BCI systems.
• This work contributes to more reliable and effective BCI applications through improved EEG analysis.