Do angry musicians play better? Measuring emotions of jazz musicians through body sensors and facial emotion detection

Using sensor data from a four-hour jazz rehearsal with 30 musicians we predicted their emotions. Ten musicians were equipped with smartwatches to collect their body signals, while a camera was recording the emotions shown in their faces. We compared the emotions calculated by the physiological signals to those based on facial features. Moreover, the relationships between emotions and physical indicators, like heart rate and speech volume, were investigated using machine learning models that predicted the intensity of facial emotions using physiological signals and emotions as inputs. We find that facial expressions of anger of the musicians go together with stress and happiness measured through the smartwatch, indicating attaining a possible state of flow.