Practical Suitability of Emotion Recognition from Physiological Signals by Mainstream Smartwatches

In the paper, we analyze the current opportunities and limitations of emotion recognition in real-life situations via mainstream smartwatches (e.g. AppleWatchT). We have identified and taken into account specific real-life situations capable to be recognized by a smartwatch app, where emotion articulation will be superimposed by physiological reactions of the human body. If not handled, such situation would result in misinterpreted emotions. Unfortunately, only one dimension of emotion, tension resp. stress, today can be securely recognized by mainstream smartwatches and only for more strong emotion articulations. To pave the way for the recognition of the other motion dimensions, arousal and valence, we propose a new test scenario, watching soccer games, as an internationally useable, highly scalable and extensively automatable test field. Only with broader experiments in this proposed field the targeted progress in emotion recognition by mainstream smartwatches will be achievable.