EEG Based Epileptic Seizure Detection Using Empirical Mode Decomposition and Hidden Markov Model Dash Deba Prasad1, Kolekar H Maheshkumar2 1Scholar, Electrical Engineering Dept., Indian Institute of Technology Patna, India 2Associate Professor, Electrical Engineering Dept., Indian Institute of Technology Patna, India Online published on 29 December, 2017. Abstract Epilepsy is a chronic neurological disorder which is indicated by recurrent seizure. According to World Health Organization about 50 million people worldwide and 80% people with epilepsy belongs to low or middle income group. Two million new epilepsy cases occur each year globally as estimated by world health organization. Present method of seizure detection is manual making the process time taking and doctor dependent. The proposed algorithm automatically detects seizures with higher accuracy. Hidden Markov model (HMM) based classification approach is proposed for epileptic seizure detection. Electroencephalogram (EEG) signal was decomposed using empirical mode decomposition. Higuchi's fractal dimension and Shannon, collision, minimum entropy features were extracted from six intrinsic mode function and average feature values were used for classification. Features extracted from the signals were efficient in differentiating seizure, healthy and inter-seizure EEG signals. K means clustering algorithm was used for generating symbol sequence. Baum-Welch algorithm was used training HMM model. Viterbi algorithm was used to find the state sequence for each observed sequence obtained after manual clustering of test signal features. Maximum accuracy of 99.16% was observed for healthy-seizure, 95.00% for seizure-Interseizure and 50.62% for healthy-Interseizure EEG signals classification. Top Keywords Epilepsy, EEG, Empirical Mode Decomposition, HMM, Least Square Support Vector Machine. Top |