Filter Bank Common Spatial Patterns in Mental Workload Estimation

TL;DR

This paper introduces a filter bank common spatial patterns algorithm combined with feature selection to improve EEG-based mental workload estimation, especially with low-cost sensors, demonstrating superior classification accuracy.

Contribution

The paper presents a novel spatial filtering method that enhances feature extraction for mental workload estimation using low-cost EEG devices.

Findings

Proposed algorithm outperforms existing methods in classification accuracy.

Effective in discriminating different mental workload levels.

Validated with data from low-cost EEG headset.

Abstract

EEG-based workload estimation technology provides a real time means of assessing mental workload. Such technology can effectively enhance the performance of the human-machine interaction and the learning process. When designing workload estimation algorithms, a crucial signal processing component is the feature extraction step. Despite several studies on this field, the spatial properties of the EEG signals were mostly neglected. Since EEG inherently has a poor spacial resolution, features extracted individually from each EEG channel may not be sufficiently efficient. This problem becomes more pronounced when we use low-cost but convenient EEG sensors with limited stability which is the case in practical scenarios. To address this issue, in this paper, we introduce a filter bank common spatial patterns algorithm combined with a feature selection method to extract spatio-spectral features discriminating different mental workload levels. To evaluate the proposed algorithm, we carry out a comparative analysis between two representative types of working memory tasks using data recorded from an Emotiv EPOC headset which is a mobile low-cost EEG recording device. The experimental results showed that the proposed spatial filtering algorithm outperformed the state-of-the algorithms in terms of the classification accuracy.