Linda Kosanke’s research while affiliated with Technische Universität Berlin and other places

The mixing time of room impulse responses denotes the moment when the diffuse reverberation tail begins. A diffuse ("mixed") sound field can physically be defined by (1) equidistribution of acoustical energy and (2) a uniform acoustical energy flux over the complete solid angle. Accordingly, the perceptual mixing time could be regarded as the moment when the diffuse tail cannot be distinguished from that of any other position or listener's orientation in the room. This, for instance, provides an opportunity for reducing the part of binaural room impulse responses that has to be updated dynamically in Virtual Acoustic Environments. Several authors proposed model- and signal-based estimators for the mixing time in rooms. Our study aims at an evaluation of all measures as predictors of a perceptual mixing time. Therefore, we collected binaural impulse response data sets with an adjustable head and torso simulator for a representative sample of rectangular shaped rooms. Altering the transition time into a homogeneous diffuse tail in real time in an adaptive, forced-choice listening test, we determined just audible perceptual mixing times.We evaluated the performance of all potential predictors by linear regression and finally obtained formulae to estimate the perceptual mixing time from measured impulse responses or physical properties of the room.

The mixing time of room impulse responses denotes the moment when the diffuse reverberation tail begins. A diffuse sound field can physically be defined by 1) equidistribution of acoustical energy and 2) a uniform acoustical energy flux over the complete solid angle. Accordingly, the perceptual mixing time is the moment when the diffuse tail cannot be distinguished from that of any other position in the room. This provides an opportunity for reducing the length of binaural impulse responses that are dynamically exchanged in virtual acoustic environments (VAEs). Numerous model parameters and empirical features for the prediction of perceptual mixing time in rooms have been proposed. This study aims at a perceptual evaluation of all potential estimators. Therefore, binaural impulse response data sets were collected with an adjustable head and torso simulator for a representative sample of rectangularly shaped rooms. Prediction performance was evaluated by linear regression using results of a listening test where mixing times could be adaptively altered in real time to determine a just audible transition time into a homogeneous diffuse tail. Regression formulae for the perceptual mixing time are presented, conveniently predicting perceptive mixing times to be used in the context of VAEs.