Pleasure and Displeasure Identification from fNIRS Signals

This paper presents two feature extraction methods for training different classification models for the detection of pleasure and displeasure defined by high and low valence levels using functional near-infrared spectroscopy (fNIRS). The study involved fifty-four volunteers who were presented with emotion-inducing image blocks while their prefrontal cortex brain activity was recorded by an fNIRS device. Results from the participants’ responses to a questionnaire showed no significant differences in valence-related scores. The study used statistical and ROCKET methods to extract features and trained six models to discriminate high and low valence. The results showed that ROCKET features performed best with kNN, MLP and SVM classifiers. The research highlighted the potential of different feature extraction methods to improve the accuracy of biosignal analysis from fNIRS devices.