Subject Independent Epileptic Seizure Recognition Using 1D-EEGNet Model

Epileptic seizures are sudden and unpredictable changes in brain activity that can significantly impact a patient's health and quality of life. While electroencephalography (EEG) is the main tool for diagnosing seizures, interpreting long-term recordings manually is still time-consuming, subjective, and not practical for continuous monitoring. This study presents a compact and efficient one-dimensional EEGNet (1D-EEGNet) model that automatically detects seizures from EEG signals. The model uses depthwise and separable convolutions to capture important temporal and spatial relationships across channels. It keeps the number of parameters low, which makes it suitable for real-time use. Experiments on the Siena Scalp EEG dataset show impressive results. The model achieves 99.83% accuracy under standard evaluation and 95.35% accuracy with leave-one-subject-out (LOSO) cross-validation. A comparison shows that the proposed model consistently outperforms the 1D-CNN and 1D-ResNet baselines. These results demonstrate the power, effectiveness , and possibilities of 1D-EEGNet for real-time, wearable seizure monitoring systems.