Optimizing Stress Detection: Harnessing MobileNet-V2 with Azimuthal EEG Mapping

This investigation meticulously assesses stress classification models using benchmark datasets (D1, D2, D3). The study delves into the effectiveness of models, notably MobileNet-V2, incorporating an innovative Azimuthal Projection approach. MobileNet-V2 consistently outshines its counterparts, showcasing impressive accuracy, specificity, sensitivity, and F1 Score. Noteworthy achievements include 95.5% accuracy, 98.8% specificity, and 95.5% sensitivity on D1, with a mere 3.5-hour training time. In D2 and D3, MobileNet-V2 maintains excellence with 96% accuracy, 97.5% specificity, and 96% sensitivity, emphasizing its computational efficiency with training times of 3.8 hours on D2 and 3.6 hours on D3. These findings position MobileNet-V2, integrated with Azimuthal Projection, as the preferred choice for stress classification, effectively balancing performance and resource utilization.