<![CDATA[Detecting epileptic seizures poses significant challenges due to the complex and variable nature of EEG signals, particularly when aiming for implementation in wearable devices. The use of 64-channel EEG electrodes, while comprehensive, is impractical for wearable applications due to their size, cost, and the high computational load required for processing. The use of a single-channel EEG wearable device offers notable advantages, including reduced size and cost, making it more practical and comfortable for continuous monitoring in daily life. Additionally, the lower computational load enhances battery life and allows for real-time data processing, which is critical for timely seizure detection and intervention. This research investigates the detection of epileptic seizures using various machine learning algorithms and the power spectrum feature extraction method from EEG signals, aiming for application in wearable devices with a single-channel electrode. The study applied random forest (RF), K-nearest neighbor (KNN), decision tree (DF), support vector mechine (SVM), and logistic regression algorithms to assess their effectiveness. Results revealed that the power spectrum extraction method notably improved seizure detection accuracy, with RF and KNN achieving 93% and 92% accuracy respectively when using all EEG channels. When limited to a single channel, SVM demonstrated the highest accuracy of 82% with channel 3. These findings underscore the efficacy of the power spectrum method for EEG signal processing, providing significant improvements in accuracy and computational efficiency. The study concludes that the proposed approach is promising for enhancing epileptic seizure detection, suggesting further optimization for real-time application in wearable devices to develop accurate and efficient diagnostic tools.]]>
