Progressive Perceptual Audio Rendering for Large Multi-source Scenes

Despite recent advances, including hierarchical and sound clustering techniques, perceptual audio rendering technique remains a high challenge for complex virtual scenes with thousands of sound sources due to its high computation cost. In the paper, we first present a spatial-frequency measurement based approach to generate recursive cluster for multiple sound source. This method can handle more than two times the number of sources on consumer 3D audio hardware, with minimal decrease in audio quality. The perceptual pilot was then studied which indicates its necessity to increase the number of different clusters of visible sound source. We present a sound categories based cost metric using this indication. The proposed method preforms well for rendering thousands of sound sources on a workstation.