A novel approach for identifying epileptic signals using signal processing

Abstract

Abstract Electroencephalography is a technical method known to analyses the behavior of human brain. The record of neurological activity is highly effective when neuronal diseases like epilepsy are to be studied. Although, the biochemical or neuro-physical causes of epilepsy remain unknown but its detection and diagnosis is no more a mystery because of EEG records. The precise and timely diagnosis of disease prevents a lot of trouble. It is challenging to handle large and bulky EEG records. This research work aims at developing an automated method to differentiate the EEG records during an epileptic seizure from the normal EEG records. The wavelet transform has an ability to zoom in the signal to is most resourceful parts and thus remove any ambiguous information from the signal. Further, the generation of useful feature set helps in setting up differentiation criteria between two types of signals. Next, the classification using neural network and support vector machine completes the process to identify the diseased signals. Also, this study contemplates the dissimilarity between two classifying techniques in order to determine the most efficient method for rational classification. The method proves to be very efficient on the pre-recorded data. The data has been acquired from universally available source mentioned in the references. The work can be extended further on simultaneous monitoring of EEG records. Also, this work can be applied to determine other neurological disorders, such as schizophrenia.