ندا كلانترمحمداباد

Epilepsy is one of the most common and devastating neurological disorders that causes recurrent seizures and is usually diagnosed by visual inspection of electroencephalogram (EEG) signals that reflect the electrical activity of the human brain. However, visual inspection of vast amount of EEG data is very time consuming and requires expertise. Therefore, the development of an automatic epileptic seizure detection system is very valuable and helps to improve the lives of patients with epilepsy. In this thesis, an automated classification method for automatic seizure detection based on signal modeling is proposed, which includes three main steps: In the first step, the EEG signals are decomposed into delta, theta, alpha, beta, and gamma brain rhythms by a multi-rate filter bank that is constructed using the basis vectors of discrete cosine transform (DCT). Then, in the next step, these brain rhythms are statistically modeled with a class of self-similar Gaussian random processes called fractional Brownian motion (fBm) and fractional Gaussian noise (fGn), the self-similarity index of these processes is called Hurst exponent (H). Finally, the features extracted from each brain rhythm include Hurst exponent and autoregressive moving average (ARMA) model parameters are used to design a support vector machine classifier and a neural network classifier for classifying EEG signals in three different modes: pre-ictal, ictal and inter-ictal. eva‎luation of classification performance is performed using two widely used EEG data set with 10-fold cross-validation method. The results of simulations (in MATLAB software environment) show that the proposed method is robust and provides good classification accuracy in both data set. Also, the performance of the proposed method is investigated in the presence of simulated noise and artifacts and the results show promising performance in the presence of noise and artifacts that exist in practical applications. Therefore, the proposed method also has high robustness against noise sources.

سيگنال EEG , بانك فيلتر با بردارهاي پايه DCT , تشخيص تشنج صرعي , فرآيندهاي تصادفي گاوسي خود-متشابه

EEG signal , Filter bank with basis vector of DCT , Detection of epileptic seizure , self-similar Gaussian random processes

neda kalantar

javad poshtan