ABSTRACT
The electroen-cephalogram (EEG) is a collection of recorded signals that represent the electrical activity in the brain. Signal processing is a way in which the EEG record can be converted into a numerical description of features in the data. The assumption of weak stationarity over windows of 20-30 seconds in length of EEG data underpins most signal processing tools used to extract features from EEG sequences. Power spectral density (PSD) analysis is an important tool used to understand the static and dynamic properties of the EEG, where static properties refer to locally stationary behavior, and dynamic properties aim to capture the time-evolving nature of the EEG. Wavelet filters isolate spectral information in different frequency ranges and can describe the EEG in similar ways as the PSD. Designing a classifier for a complex system such as the epileptic EEG involves several stages of signal processing.
