EEG SIGNAL PROCESSING: A Machine Learning Based Framework

1.1 Motivation Analysis of non-stationary and non-linear nature of signal data is the prime talk in signal processing domain today. On employing biomedical equipments huge volume of physiological data is acquired for analysis and diagnostic purposes. Inferring certain decisions from these signals by manual observation is quite tedious due to artefacts and its time series nature. As large volume of data involved in biomedical signal processing, adopting suitable computational methods is important for analysis. Data Science provides space for processing these signals through machine learning approaches. Many more biomedical signal processing implementations are in place using machine learning methods. This is the inspiration in adopting machine learning approach for analysing EEG signal data for epileptic seizure detection. Epilepsy is a chronic neurological disorder and 0.7% of world population is affected by this disorder according to WHO. Though many researchers from medical and computer science fields have contributed their knowledge for developing automated diagnosis systems (also called Diagnosis Decision Support System) by using scientific techniques and authenticated EEG data bases, the efficiency of these systems are still questionable due to several reasons. For the last one decade many researchers have been bringing automated systems for epileptic seizure detection using machine learning methods. As a continuous effort to enhance the accuracy in epileptic seizure detection, this machine learning framework is brought. Few knowledge areas which have contributed to this research are being introduced in the following sections.