Comparison of time-frequency feature extraction methods for EEG signals classification

Abstract

Biomedical signal analysis can assist the diagnostic process for each neurological dysfunction. At that time the diagnostic tool continues to evolve, which is related to the development of both technological and innovative solutions proposed by scientists. One of possible solutions for automatic EEG analysis is to use modern signal processing tools, which are able to give a time-frequency representation of a signal. This paper presents a comparison of methods such as the discrete wavelet transform, the matching pursuit algorithm, and the S transform for feature extraction and then classification of neurological disorders. The research was carried out using real EEG recordings of epileptic patients as well as healthy subjects. © 2013 Springer-Verlag.