Improved epileptic seizure detection using singular spectrum empirical mode decomposition and machine learning approach

Electroencephalography (EEG) is multi-channel electrical signal acquiring tool used to acquire the brain signals and used in various application such as neurological disorder detection and prediction, monitoring psychological condition, emotion analysis and Brain-Computer Interface (BCI). Epilepsy is a highly prevalent neurological disorder caused by the occurrence of seizures. Epilepsy is a neurological condition of the brain where some neurons change its behaviours such as hyper-activeness and channel synchronization. The EEG signals recorded from epileptic patients are analysed for monitoring and extracting behaviour of signals during onset seizures. The Time and Frequency Domain (TFD), Wavelet Transform (WT) and Empirical Mode Decomposition (EMD) are well-proven feature extraction methods used for various applications. The objective of the paper is to propose a new effective method, Singular Spectrum Empirical Mode Decomposition (SSEMD) for effective classification of Normal and Epileptic EEG Signals. The high-performance machine learning classifiers are used for classification of EEG signal in normal and epileptic class. The performance observed with the proposed feature extraction method is 99.8 percent of detection accuracy with nearly zero false positive rates. The average dimensionality reduction of 70 percent of total feature space is observed due to the use of ANOVA.