小型fNIRS装置を用いた総ヘモグロビンデータによる快・不快情動の推定

抄録

A method for estimating human emotions using a lightweight fNIRS device is proposed. In the proposed method, from only 2ch sensors total-hemoglobin data measured by the lightweight fNIRS device, 35 features for estimating emotions are generated. Using generated features, K-Nearest Neighbor model for estimating pleasant / unpleasant emotions are constructed. The evaluation of the constructed models using measured 320 data from 20 subjects shows that it is possible to estimate the emotions with more than 80% recall by cross-validation.