Accurate depression detection is essential for timely intervention and treatment, current models leverage the rich spatial and temporal information embedded in EEG signals captured from the subjects. However, EEG signals are dynamic and needs large feature space. This oversight has significant implications for mental health diagnostics and patient outcomes, underscoring the importance of developing more effective computational models. Current deep learning models for depression detection using EEG signals have some limitations and these include the mutual exclusiveness of the temporal and spatial convolutions making it unable to rely solely on single feature extraction methods for effectively capturing of the patio-temporal characteristics of EEG data. Also, the EEG signals are highly dense and contain multiple channels, capturing such variations over multiple channels and over time is a challenge. Further, previous research failed to consider the multiple channel relationships of EEG signals at different times collectively. For the purpose of effective accuracy and precision, This study embraced the potential capability of Residue Number System (RNS) model for accuracy, sensitivity, and specificity to enhanced EEG based depression detection speed and accuracy network using Graph Convolutional Gated Recurrent Unit develop depression detection model that leverages the aforementioned limitations and methodological components for improved diagnostic accuracy The model achieves the accuracy came out to be 92.25%, the F1-score was 0.9266 and the sensitivity was 0.9483.
Keywords: electroencephalogram (EEG) signals, notch filtering, residue number system (rns) model brain waves
