360 Videos with Audio Skew

In this study, we explored the impact on Quality of Experience (QoE) of different audio skews in 360{\degree} videos. Whilst also collecting subjective QoE responses, we focused mainly on objective QoE metrics, as provided by physiological readings from Heart Rate and Galvanic Skin Response sensors. To this end, we presented three different 360{\degree} videos, each with distinct motion dynamism and audio skew, to a group of 30 participants. The physiological data collected was analyzed using a variety of machine learning (ML) models which achieved high accuracy in predicting objective user QoE. Our results also revealed that higher audio skews led to an increased level of stress, as evidenced by more pronounced physiological signals. This suggests that audio dynamics can significantly impact the user experience in 360{\degree} videos, as higher audio dynamism was more susceptible to audio skews. Lastly, the high performance of our models, particularly those based on Decision Tree (DT)  and Random Forest (RF) approaches highlights the potential of ML, notwithstanding challenges such as data high dimensionality and heterogeneity, to build objective QoE models in 360{\degree} environments.