Regresar

EEG-Based BCI Emotion Recognition Using the Stock-Emotion Dataset

Abstract:

We present a novel method of emotion elicitation through stock trad- ing activities in a competitive market for automatic emotion recognition. We pre- pared eight participants, who placed buy and sell orders of a specific and very volatile security in simulated trading with real-time market data. A “carrot and stick” approach was implemented to elicit emotions efficiently, which added risks and rewards to participants’ trading decisions. Consequently, key trading emotions were triggered: 1) fear, 2) sorrow, 3) hope, and 4) relaxed (or “focused,” which is the optimal emotional state). We gathered our data (Stock_Emotion) through an EEG-Based BCI device. Such a dataset was pre-processed, features were extracted, and kNN, MLP, and Random Forest algorithms were applied for emotion classification in the valence-arousal space. Our accuracy results were satisfactory.