GMM better than SRC for classifying epilepsy risk levels from EEG signals

EEG is the most commonly used technique for the monitoring of the brain activities in the case of epilepsy. Generally, the visual encephalographers and the expert neurologists analyze the EEG records and it is quite time consuming. In the EEG recordings, several noisy characteristics are present making it difficult for the experts to differentiate between an artifact and seizure in raw EEG signals. The detection and classification of epileptic activity is quite a high demanding process. Therefore automatic prediction and detection algorithms are in need to predict the epileptic seizures in the EEG signals. The technique used here is by the employment of Approximate Entropy as a Feature Extraction technique and then Gaussian Mixture Model (GMM) and Sparse Representation Classifier (SRC) are used as Post Classifiers for the classification of epilepsy risk levels from EEG signals. The bench mark parameters analyzed here are Performance Index (PI), Quality Value (QV), Specificity, Sensitivity, Time Delay (TD) and Accuracy.