Decoding of Binary Mental Arithmetic Based Near Infrared Spectroscopy Signals

There has been an increase interest for functional near-infrared spectroscopy (MRS) in recent years since it is a non-invasive technique as well as few restrictions to the subjects and not affected by electrical noise. In this study, we analyzed mental arithmetic based NIRS signals that it can he helpful for patients like dyscalculia where difficulty learning or lack of attention problem exists. So, it is important that the mental arithmetic is effectively separated from NIRS signal. For this purpose, first, we determined change in oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HbR) concentrations by applying the modified Beer-Lambert law to NIRS data set. After Hilbert transform (HT) + sum derivative (SD) based features were extracted from pre-processed HbR and HbO, these features were classified by k-nearest neighbors. The average classification accuracy (CA) rates of 82.87% and 84.94% were calculated from the HT+SD based features that best determine the mental arithmetic of the HbR and HbO signals, respectively. It can be said that the proposed method is effective for this dataset, in view of the fact that these values are 2.17% and 1.34% higher than CAs calculated in the literature for HbR and HbO, respectively.