A More Complete Picture of Emotion Using Electrocardiogram and Electrodermal Activity to Complement Cognitive Data

摘要

We describe a method of achieving emotion classification using ECG and EDA data. There have been many studies conducted on usage of heart rate and EDA data to quantify the arousal level of a user. Researchers have identified a connection between a person's ECG data and the positivity or negativity of their emotional state. The goal of this work is to extend this idea to human computer interaction domain. We will explore whether the valence/arousal level of a subject's response to computer based stimuli is predictable using ECG and EDA, and whether or not that information can complement recordings of participants' cognitive data to form a more accurate depiction of emotional state. The experiment consists of presenting three types of stimuli, both interactive and noninteractive, to 9 subjects and recording their physiological response via ECG and EDA data as well as fNIRS device. The stimuli were selected from validated methods of inducing emotion including DEAP dataset, Multi Attribute Task Battery and Tetris video game. The participants' responses were captured using Self-Assessment Manikin surveys which were used as the ground truth labels. The resulting data was analyzed using Machine Learning. The results provide new avenues of research in combining physiological data to classify emotion.