Deep Learning Based EEG Based Mental Workload Detection with Discrete Wavelet Transform and Welch’s Power Spectral Density

EEG data provide a convenient and portable method of measuring mental workload. This work analyzes EEG data collected during rest intervals and arithmetic problems using deep learning and machine learning approaches. EEG signals were broken down into sub-bands using the Discrete Wavelet Transform (DWT), and the power levels of these bands were estimated using Welch’s approach. Based on these power levels, feature vectors were identified using both conventional machine learning techniques including Linear Discriminant Analysis (LDA), k-Nearest Neighbors (k-NN), and Support Vector Machines (SVM) and deep learning algorithms like Long Short-Term Memory (LSTM). The findings show that LSTM performed better than conventional classifiers, with higher performance metrics and accuracy. This demonstrates how deep learning may be used to separate mental workload states from EEG signals with accuracy.