Time-Frequency EEG and Bi-LSTM-Based Automated Cognitive Load Detection Using Machine Learning

In the objective study, a new way of detecting cognitive strain is detailed utilizing 1D-EEG signals and their various time-frequency (T-F) representation as 2D pictures. A Bi-LSTM framework incorporating several optimization techniques was used for categorization after the data was normalized and artifact-free segmented using filter-based methods. Training and fine-tuning of the Bi-LSTM networks was completed. Individualized cognitive functions may be better understood with the help of this work’s subject-wise mental load detecting them. Psychological surveillance in several domains and focused therapies are both made possible by this strategy. In numerous diagnosis scenarios, EEG signal processing relies on removing EOG artifacts. Understanding the brain’s response to stimuli requires investigations into brain function. This thesis details novel methods for removing EOG artifacts and determining mental strain using EEG data. A few other cognition-related objectives still need the attention of scientists. A wide range of occupations need for different sets of abilities, such as paying attention, retaining information, recall, and judgement. The proposed Bidirectional Long Short-Term Memory (Bi-LSTM) model achieves a flawless accuracy rate of 100%, surpassing earlier models such as Stacked-LSTM, CNN-LSTM, LSTM-RNN, EEGnet +CNN, and CNN + LSTM. The Bi-LSTM model nevertheless outperforms the competition when it comes to identifying cerebral stress, even when these other methods are successful.