CLASSIFICATION OF FUNCTIONAL NEAR-INFRARED IMAGING BASED HEMODYNAMIC PATTERNS RECORDED AT MENTAL ARITHMETİC AND RESTING

Abstract

Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical imaging technique used in brain-computer interface (BCI) systems. It is used to measure deoxyhemoglobin and oxyhemoglobin proportions that occur during a specific activity in the brain region (motor and visual activity, auditory stimulus, etc.). In this study, hemodynamic patterns were recorded from 8 participants during mental arithmetic and rest activities. Features have been extracted for this by using detrended fluctuation analysis, entropy and Hjorth parameters methods. The distinctive feature vectors obtained after the feature selection process have been applied to support vector machines (SVM), multilayer artificial neural networks (MLANN) and k-nearest neighbors (k-NN) classifiers. As a result, the best classification accuracy was 97.17% when SVM classifier was used.