Akademik Veri Yönetim Sistemi
JOURNAL OF POLYTECHNIC-POLITEKNIK DERGISI, cilt.28, sa.5, 2025 (ESCI, TRDizin)
Epilepsy is a neurological disorder that affects millions of people worldwide, and the rapid and accurate detection of epileptic seizures is crucial in improving patients' quality of life. This study performs various analyses using different power spectral density methods and classifiers for epileptic seizure detection from EEG signals. Methods such as Music, Lomb-Scargle, Multitaper, Welch, Periodogram, and Burg are tested to identify changes in their ability to distinguish spectral resolution and frequency components. Reference signals are created for each class, and discriminative features such as spectral energy, spectral entropy, and maximum spectral deviation are extracted by comparing these reference signals. These feature vectors are used in classification with Random Forest and k-Nearest Neighbor algorithms. The results indicate that the high-resolution spectral power density methods, Music and Lomb-Scargle, along with the Random Forest classifier, achieved the highest accuracy. This study makes a significant contribution to the literature by demonstrating that the combined use of high-resolution spectral power density methods and powerful ensemble learning-based classifiers can significantly improve seizure detection accuracy.