| Summary: | 碩士 === 國立成功大學 === 電機工程學系碩博士班 === 97 === Epilepsy is one of the most common neurological disorders, and approximately 1% of people in the world suffer from epilepsy. Epilepsy is caused by abnormal discharges in the brain, thus electroencephalogram (EEG) signal has been an especially valuable clinical tool for the detection and diagnosis of epilepsy. An expert detects epileptic activity by visual inspection of the EEG, which is a time-consuming procedure for recordings that are days long. In addition, the subjective nature of the examination affects the outcome. Hence, automation of this process could save time, making the decision more reliable.
In this paper, we proposed an architecture for classification problem in five-class epileptic EEG-signals. We tried to classify EEG signals with support vector machine (SVM) according to features which were made by discrete wavelet transform (DWT) and approximate entropy (ApEn). The EEG signals were decomposed into five subbands by using DWT, then the ApEn values in subbands were computed. The feature vectors for classification were selected adaptively from statistical wavelet coefficients and ApEn values by proposed feature selection method. Finally, the SVMs were used for classifying the selected features. The experimental results showed the proposed system has great performance and reliability, and the total accuracy of classification could achieve 98%.
|
|---|
