Emotion Recognition by Heart Rate Variability

Background: Emotion plays an important role when people face difficult social problems in their daily activities. This study explores the application of sensors and mobile technologies to detect and recognize school bullying. Many databases offer data for emotion recognition research. One of these is the Mahnob-HCI-Tagging database, which yields a baseline accuracy for emotion recognition based on EEG, eye gaze, and a combination of EEG and eye gaze. Because EEG and eye gaze are not suitable for emotion recognition in the mobile gadget environment, it is interesting to investigate other physiological signals individually, such as ECG, galvanic skin conductance (GSR), body temperature and respiration rate. Objective: This paper focussed on the ECG signal and, more specifically, on heart rate variability (HRV), derived from ECG, to identify certain standard features used in emotion recognition. Instead of using discrete emotions as labels, we transferred emotions, such as fear, anger, happiness and anxiety, to an arousal-valence space. Results:For arousal and valence based on HRV, the baselines are 47.69 and 42.55, respectively, while those for arousal and valence in all physiological signals were 46.2 and 45.5, respectively. The most challenging label in this experiment turned out to be #65533neutral#65533 in the valence scale, as the SVM classified all results as either #65533unpleasant#65533 or #65533pleasant#65533. Conclusion: This work provided a baseline for emotion recognition research based on ECG signals. It also encourages experimental trials using GSR, body temperature and respiration rate individually