Abstract:
The auralization of room impulse responses is generally realized by direct convolution with an anechoic signal. When using a simulated impulse response, the realism of the auralization relies mainly on the accuracy of the propagation model used to estimate time and spatial distributions of early reflections and of the reverberation process. An alternative approach, adopted at Ircam, uses a room effect synthesizer driven by acoustical indices, derived from the analysis of the response in order to describe the corresponding spatial, energetic, and statistical properties. Although the auralization performance then depends on the relevancy of these parameters, this latter solution offers the following advantages: (1) reduction of the computational cost; (2) manipulation of measured, estimated, and virtual configurations in a unified perceptual formalism; (3) adaptation to different reproduction format or setup, without recomputing the corresponding impulse responses. In order to validate the performance of the synthesis approach, a series of comparisons between synthesized and measured responses is undertaken, focusing successively on different time sections of the response. The study of the late reverberation is conducted by comparing measured monaural or binaural responses with modified responses where the late section is replaced by a decreasing white noise, or by a reverberation signal provided by a feedback delay network. These modifications are controlled by analysis/synthesis methods based on time-frequency representations.