Feature extraction of EEG Signal based on Savitzky-Golay filter

Abstract

EEG is a technique that enables the extraction and processing of brain signals. This study is important to observe the complex behaviour of the brain signals in order to categorize the signals based on the patterns. This plays a pivotal role in early diagnosis and prediction of brain disorders. The objective of this paper is to design a system to analyze EEG signals. This system will be useful in real time and medical diagnostic environments. The foremost stage is filter designing. It is a crucial step as the obtained brain signals are infiltrated by noise. The filter should be designed such that the noise is removed without affecting the quality of the signal. Initially in this work, synthetic EEG signals are tested on various filters. This is followed by extracting features from the filtered signal. By observing the values of these extracted features, a decision is made as to which filter is best suited for the study of EEG signal. It is found that Savitzky-Golay (SG) filter yields the best results. Following this, normal EEG signal and epileptic EEG signal is filtered by SG filter followed by feature extraction. Various features such as energy, entropy and band power are determined for both the classes.