HCC: Small: Making Virtual Reality Safe

尽管消费者级别的虚拟现实头戴式显示器已经变得足够便宜，可以让广大用户购买，但仍然存在严重的安全问题。头戴式显示器遮挡了周围的现实世界，这可能会隐藏障碍物，如桌子、宠物、墙壁或其他潜在的碰撞危险。当前避免碰撞的方法依赖于用户定义游戏区域边界；此过程可能因用户错误而导致人身伤害。此外，目前的方法对于需要快速动作的游戏来说是无效的，因为系统可能无法及时做出反应以防止受伤。为了解决这些问题，研究人员将开发基于摄像头的方法，用于实时检测潜在碰撞，并评估反馈技术，以减少使用虚拟现实耳机时的伤害发生率。这种方法有可能使虚拟现实在现实环境中的使用更加安全。该项目的目标是创建和评估重建、分割和运动预测技术，以提供避障反馈，并减少使用虚拟现实头戴式显示器的人员受伤的发生率。与之前的方法相比，根据用户身体靠近真实障碍物的特定位置提供量身定制的反馈，理想情况下可以减少用户与现实世界物体的碰撞。具体而言，该团队将1)确定真实障碍物检测和分割的最佳方法；2)研究全身运动预测方法的有效性；3)确定真实障碍物警报的最佳模式和位置，以最大限度地提高存在感，同时最大限度地减少碰撞；4)评估真实障碍物警报系统在虚拟现实系统中的纵向影响。最终，这项研究将产生人类运动数据集，可用于未来虚拟现实研究中的运动预测，以及开源插件，通过减少伤害，使当前的虚拟现实体验更安全。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Although consumer-level virtual reality head-mounted displays have become affordable enough for a broad user base to purchase, there are still serious concerns about safety. Head-mounted displays block out the surrounding real world, which can hide obstacles, such as tables, pets, walls, or other potential collision hazards. Current approaches to avoiding collisions depend on the user to define play area boundaries; this process is subject to user error and thus can lead to injury. Moreover, current approaches are ineffective for games that require fast motions, as the systems may not react in time to prevent injury. To address these problems, the investigators will create camera-based methods for detecting potential collisions in real time and evaluate feedback techniques to reduce the incidence of injury when using virtual reality headsets. This approach has the potential to make the use of virtual reality much safer in real-world environments.The objective of this project is to create and evaluate reconstruction, segmentation, and motion-prediction techniques to inform obstacle avoidance feedback and reduce the incidence of injury to people using virtual reality head mounted displays. Providing feedback tailored to the specific locations of the user's body that are in proximity to real obstacles will, ideally, reduce user collisions with real-world objects compared to prior approaches. Specifically, the team will 1) determine the best approaches to real obstacle detection and segmentation, 2) investigate the efficacy of full body motion prediction approaches, 3) ascertain the optimal modality and locations of real obstacle alerts to maximize presence while minimizing collisions, and 4) evaluate the longitudinal impact of real obstacle alert systems in virtual reality systems. Ultimately, this research will result in human motion datasets that can be used for future motion prediction in virtual reality research, as well as open-source plugins that will make current virtual reality experiences safer by reducing injuries.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.