Acoustic Rendering for Virtual Environments

In order for auralization to better replicate the binaural listening experience in a space, acoustic scattering in the room impulse response (RIR) must be more accurately calculated. Three models are discussed and used in room computations and auralizations: Lambert scattering, edge-diffraction models based on the Biot-Tolstoy-Medwin technique, and boss models for protuberances on reflecting surfaces. A psychoacoustic study with the Lambert “diffusion” model shows that the ear can clearly hear frequency-dependent changes in the scattering coefficient from 125 – 4000 Hz and that the perceived quality of the changes depends greatly on the temporal and spectral characteristics of the input signal. Edge-diffraction modeling offers a greater degree of accuracy compared to Lambert-diffusion, and a validated time-domain model is applied to calculate the RIR for a stage house, showing that reflected-diffracted combinations are significant and that even small spectral changes of up to 2 dB are audible as coloration differences in the room response. Finally, a boss model of scattering from baffled hemispheres is implemented, and binaural impulse responses are used in initial listening tests to investigate the audibility of the boss scattering. As an improvement upon Lambert-diffusion methods, a hybrid model is proposed incorporating BTM-based edge-diffraction and boss models.