Daily Emotion Recognition from Physiological and Environmental Data using Wearable Devices

Daily emotion recognition (DER) faces difficulties due to different environments, especially outdoor settings. We envision that DER will enhance various applications. For example, a smart healthcare system could detect emotions in both indoor and outdoor settings. This study proposed a real-time monitoring method named envER, which integrates environmental data with physiological data from wearable devices. A comprehensive validation of envER was conducted, applying three feature selection methods, and two model validation techniques to the publicly available DAPPER dataset. All assessments highlight the importance of environmental data in emotion recognition, and when combining environmental data with body data consistently yields improved performance in both binary classifications of arousal and valence. Finally, envER can achieve the highest recognition accuracies of 87.2% for arousal and 95.5% for valence.