Spatial Impulse Response Rendering II: Reproduction of Diffuse Sound and Listening Tests

Spatial impulse response rendering (SIRR) is a method for reproducing room impulse responses over multichannel loudspeaker setups. The applied analysis and synthesis methods were introduced in a companion paper. Time-frequency analysis is used to obtain directional and diffuseness information from the recorded sound field. Nondiffuse sound is then reproduced as pointlike virtual sources, and diffuse sound is synthesized with a decorrelation technique. The proposed synthesis methods for diffuse sound are examined in more detail and a hybrid method is derived. The relationship between diffuseness and interaural coherence is also studied. In addition, results of two listening tests are presented. It is shown that with a large loudspeaker setup under anechoic conditions, SIRR reproduction is at best indistinguishable from the original sample. Furthermore, in a listening test conducted in a standard listening room with real measured responses, SIRR reproduction is evaluated as the most natural one of the systems studied.