Where are we?: evaluating the current rendering fidelity of mobile audio augmented reality systems

Mobile audio augmented reality systems (MAARS) simulate virtual audio sources in a physical space via headphones. While 20 years ago, these required expensive sensing and rendering equipment, the necessary technology has become widely available. Smartphones have become capable to run high-fidelity spatial audio rendering algorithms, and modern sensors can provide rich data to the rendering process. Combined, these constitute an inexpensive, powerful platform for audio augmented reality. We evaluated the practical limitations of currently available off-the-shelf hardware using a voice sample in a lab experiment. State of the art motion sensors provide multiple degrees of freedom, including pitch and roll angles instead of yaw only. Since our rendering algorithm is also capable of including this richer sensor data in terms of source elevation, we also measured its impact on sound localization. Results show that mobile audio augmented reality systems achieve the same horizontal resolution as stationary systems. We found that including pitch and roll angles did not significantly improve the users' localization performance.