IncreasingWeb3D Accessibility with Audio Captioning

Situational awareness plays a critical role in daily life, enabling individuals to comprehend their surroundings, make informed decisions, and navigate safely. However, individuals with low vision or visual impairments face difficulties in perceiving their real or virtual environment. In order to address this challenge, we propose a 3D computer vision-based accessibility solution, empowered by object-detection and text-to-speech technology. Our application describes the visual content of a Web3D scene from the user's perspective through auditory channels, thereby enhancing situational awareness for individuals with visual impairments in virtual and physical environments. We conducted a user study of 44 participants to compare a set of algorithms for specific tasks, such as Search or Summarize, and assessed the effectiveness of our captioning algorithms based on user ratings of naturalness, correctness, and satisfaction. Our study results indicate positive subjective results in accessibility for both normal and visually-impaired subjects and also distinguish significant effects between the task and the captioning algorithm.