Abstract:
Workload, or the amount of cognitive processing required to complete a task, has a big impact on productivity, security, and general well-being. A method for detecting mental workload uses electroencephalography (EEG) measurements, which document the electrical activity of the brain. To train deep learning models to categorize different levels of workload, signal processing techniques are used to extract features that directly relate to mental workload.
A crucial step in analyzing EEG signals to determine mental workload is feature extraction. The frequency distribution, variability, and complexity of EEG signals linked with various workload levels can be analyzed using statistical techniques like
mean, variance, skewness, and kurtosis. Alpha, beta, and theta band power are spectral features that were specifically designed to capture distinct patterns and correlations within the EEG signals indicative of workload levels.
The LSTM (Long Short-Term Memory) and RNN (Recurrent Neural Network) models, which are excellent at learning from the past and capturing temporal correlations in data, are fed with the retrieved statistical and hand-crafted features. The project recognizes the importance of domain expertise in the feature engineering process, even as it understands the drawbacks and difficulties associated with feature extraction approaches, such as subjective feature selection and probable information loss due to dimensionality reduction. The results showed that the LSTM and RNN models outperformed other models, obtaining a 92% machine-learning classification accuracy. Deep learning algorithms for further improvement produced an accuracy of
89%.
In conclusion, our project measures and categorizes mental workload levels using EEG signal analysis and deep learning models. The study achieves excellent accuracy in workload classification by utilizing the sequential pattern of EEG signals, extracting statistical and handmade features, and these features. These findings have important ramifications for enhancing performance, security, and general well-being across a range of sectors and disciplines where effective mental workload management is essential.
