Subjective Evaluation of Diffuse Reverberation in TASCAR (online)
* Presenting author
TASCAR is a toolbox for rendering interactive virtual acoustic environments. To increase the plausibility of these environments, it is possible to generate reverberation in two different ways: The "foaconv" plugin implements a partitioned convolution with First Order Ambisonics (FOA) impulse responses. The "simplefdn" plugin implements a simple Feedback Delay Network (FDN). Both approaches are evaluated and validated by comparing them with real recorded binaural room impulse responses for different rooms. For this purpose, objective room acoustic measures, in particular the T60 and the Direct-To-Reverberant-Ratio (DRR), are considered as well as a subject study in which perceptual measures such as the naturalness or the metallic character of the scenes are evaluated. The obtained results of the two evaluation methods are presented and interpreted, and possible implications for reverberation generation in TASCAR (or similar tools) are derived. The results show that plausible reverberation generation in TASCAR is possible with simple means. However, the room acoustic measures cannot be reliably predicted for different room geometries. Manual adjustments of processing parameters might be necessary to match the desired room acoustic properties.