Features extraction based on Schrödinger operator's spectrum for cognitive states classification

Abstract

Training machine learning algorithms to classify cognitive states is a challenge that many biomedical researchers are dealing with nowadays, for the numerous medical advantages that this kind of research has in understanding many neurodegenerative diseases. However, it is important to feed these classifiers with high-quality features allowing us to obtain high classification performance of cognitive states. We propose in this work, a new signal analysis modality to extract features from some specific brain regions whose activations are triggered by two mental states, performed by different subjects. We explore the efficiency of the technique and its fundamental aspects.