IQ estimation by means of EEG-fNIRS recordings during a logical-mathematical intelligence test

Intelligence differences of individuals are attributed to the structural and functional differences of the brain. Neural processing operations of the human brain vary according to the difficulty level of the problem and the intelligence level of individuals. In this study, we used a bimodal system consisting of functional Near-Infrared Spectroscopy (fNIRS) and Electroencephalogram (EEG) to investigate these inter-individual differences. A continuous wave 32-channel fNIRS from OxyMonfNIRS device (Artinis) and 19-channel EEG from (g.tec's company) were utilized to study the oxygenation procedure as well as the electrical activity of the brain when doing the problems of Raven's Progressive Matrix (RPM) intelligence test. We used this information to estimate the Intelligence Quotient (IQ) of the individual without performing a complete logical-mathematical intelligence test in a long-time period and examining the answers of people to the questions. After EEG preprocessing, different features including Higuchi's fractal dimension, Shannon entropy values from wavelet transform coefficients, and average power of frequency sub-bands were extracted. Clean fNIRS signals were also used to compute features such as slope, mean, variance, kurtosis, skewness, and peak. Then dimension reduction algorithms such as Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA) were applied to select an effective feature set from fNIRS and EEG in order to improve the IQ estimation process. We utilized two regression methods, i.e., Linear Regression (LR) and Support Vector Regression (SVR), to extract optimum models for the IQ determination. The best regression models based on fNIRS-EEG and fNIRS presented 3.093% and 3.690% relative error for 11 subjects, respectively.