Impacto do tipo de áudio na dispersão dos pontos de fixação em vídeos imersivos: evidências para o desenvolvimento de modelos de saliência visual

This work investigates the impact of different acoustic conditions (Ambisonics, Mono, and Mute) on visual attention in 360° videos. The analysis utilized eye-tracking data from 60 observers from the AVS-ODV dataset. The results show that the presence of audio significantly reduces fixation point dispersion, promoting greater inter-observer convergence, with no significant differences observed between Ambisonics and Mono. Our results indicate that, for the computational modeling of audiovisual saliency, spatial audio may not be strictly necessary to capture the global convergence of fixations, enabling simpler acoustic representations.