Seizure and Non-Seizure EEG Signals Detection Using 1-D Convolutional Neural Network Architecture of Deep Learning Algorithm

In this paper, seizure activities of EEG signals have been detected exploiting 1-D convolutional neural network architecture of deep learning algorithm. In past some years, many algorithms have been developed for seizure detection. Traditional method of seizure detection includes manual feature selection and classifier separately. In recent years, convolutional neural network architecture of deep learning gets very much popularity in place of traditional method. Though, very few works of convolutional neural network are found in literature for one dimensional signal classification. However, one dimensional EEG and ECG signal classifications are necessary to detect and investigate diseases like seizure activity, sleep apnea or arrhythmia and so on. The proposed method of one dimensional EEG signal classification for epileptic seizure detection involves the smallest and simplest 1-D convolutional neural network architecture for automated feature selection. The projected architecture is implemented to detect several kinds of seizure and non-seizure EEG signals and found achieving greater accuracy and sensitivity in comparison to that acquired by state-of-the-art methods involving manual feature selection and the identical EEG dataset.