An approach to real-time late reverberation generation for six-degrees-of-freedom Auditory Augmented Reality

Immersive Auditory Augmented Reality (AAR) requires the seamless integration of virtual sounds into a real environment, yet achieving this on mobile devices is challenging due to the difficulty of rendering room-matched reverberation in real-time. To address this, we previously proposed the Augmented reality Room Acoustic Estimator (ARAE) framework, which renders room-dependent spatial audio from geometry captured by on-device LiDAR sensors using a game audio engine. While this framework could dynamically generate early reflections via ray tracing, its approach for late reverberation relied on pre-generated Impulse Responses (IRs) modeling directional energy decay in a rectangular room approximation. This posed a significant limitation, as convolution with generated location-specific IRs is not well-supported by such audio platforms, thus hindering a consistent AAR experience across diverse environments. This study extends the framework by incorporating a parametric synthesizer for late reverberation, using a Directional Feedback Delay Network (DFDN). This method enables efficient parametric control of direction-dependent reverberation, using the estimated directional energy as directional parameters. This can deliver a fully adaptive AAR experience on computationally constrained mobile platforms. Future work will focus on a subjective evaluation to determine the optimal balance between perceptual plausibility and computational cost by testing listeners’ perceptual tolerance.