ECG and GSR Data for Emotion Recognition

We present the data set of Electrocardiograms (ECG) and Galvanic Skin Response (GSR) for the emotion recognition task, helpful in human-computer interaction for individuals with cognitive and physical disorders. Wearable shimmer3 ECG and GSR sensors were used to acquire data from a total of 25 participants. The dataset consists of three parts, raw data, extracted features for ECG, and self-annotation labels. Raw data consists of multimodal data and single modal data in .mat format. Multimodal data consists of ECG and GSR for 12 participants, provided with 21 stimulus videos divided into three sessions. Similarly, single modal data consists of ECG data acquired from 13 other participants. These physiological signals correlate with the six basic emotions of surprise, anger, fear, happiness, sadness, disgust, and the addition of a neutral state. Each of these seven states was further divided into five levels of intensity of felt emotion, making a total of 35 emotional states. Self annotation was performed by participants for these emotional states as well as scores of valence, arousal, and dominance. Twenty emotion-related features were extracted from each ECG sample and also provided as supplementary material for further analysis. This data provides a valuable addition to the physiological signal-based emotion recognition for the larger set of emotions and acquired during stressed situation of Covid-19 epidemic.