Parallel Mechanism of Spectral Feature-Enhanced Maps in EEG-Based Cognitive Workload Classification

Area of Science:

• Neuroscience and Brain-Computer Interfaces
• Signal Processing and Machine Learning

Background:

• Electroencephalography (EEG) offers a non-invasive, portable, and cost-effective method for monitoring neural activity.
• Classifying cognitive workload using EEG is crucial for tasks requiring high attention.
• Traditional methods like Common Spatial Pattern (CSP) and recent deep learning approaches using spectral maps have shown promise.

Purpose of the Study:

• To propose a novel parallel mechanism of spectral feature-enhanced maps for EEG-based cognitive workload classification.
• To enhance the expression of structural information in EEG signals, mitigating compression from inter- and intra-subject variability.
• To improve the accuracy of cognitive workload classification compared to existing deep neural network methods.

Main Methods:

• Utilized a public EEG dataset for evaluating the proposed method.
• Employed established deep neural networks, including AlexNet, VGGNet, ResNet, and DenseNet.
• Developed and integrated a parallel mechanism for spectral feature-enhanced maps.

Main Results:

• The proposed spectral feature-enhanced maps approach achieved a classification accuracy of 93.71% across four levels of cognitive workload.
• This represents a significant improvement over the baseline accuracy of 91.10% achieved by combining spectral maps with deep neural networks.
• The method effectively enhanced structural information, addressing challenges posed by subject variability.

Conclusions:

• The parallel mechanism of spectral feature-enhanced maps is an effective technique for improving EEG-based cognitive workload classification.
• This approach offers a more robust and accurate method for analyzing neural signals in demanding cognitive tasks.
• The findings suggest a promising direction for advancing brain-computer interfaces and attention monitoring systems.