Feature Extraction from EEG Signals: A deep learning perspective

The implementation of the learning model is primarily dependent on the features extracted from the EEG signals for any mental task classification model. A feature depicts an identifiable measurement, a unique property and a functional factor achieved from a section of a pattern. The main features which are extracted are mainly intended to limit the damage of significant data embedded in the signal and are also used to reduce resources required to illustrate a vast set of data precisely. This is important to reduce the expense of data handling, to limit the intricacy of usage and to remove the necessity of compressing the data. Different methods are applied to extract the features from EEG signals; among these methods are Empirical Wavelet Transform (EWT), Fuzzy C-means algorithm, Fuzzy Region of Interest Activity (FURIA). Studies have been made on EEG signals because of its capability to produce a method to calculate brain stimulation which is broadly used in brain-computer interface researches with uses in rehabilitation engineering and medical studies. The main objectives of this paper are to discuss some established methods of EEG feature extraction methods, comparing their performances for a particular task and suggesting the most appropriate method for feature extraction based on performance.