Experiments on Synchronous Nonlinear Features for 2-Class NIRS-Based Motor Imagery Problem

Abstract

This paper aims to experiment several synchronous nonlinear features in the well-known 2-class motor imagery problem in Brain Computer Interface (BCI) systems using Near Infrared Spectroscopy (NIRS) technique. Those features including phase synchronizations and nonlinear interdependences are well known and widely applied on several neural-related problems such as epilepsy prediction. However, only a few publications are related to NIRS-based BCI systems. We conducted several experiments using NIRS technique to analyze how useful those synchronous nonlinear features can be applied on NIRS-based BCI systems. Results show that while the nonlinear interdependences can produce quite good recall and precision ratios, the phase synchronizations are not good for classification because the accuracy is as low as that in random guessing.