Validating auralizations by using articulation indexes

In this paper, computer modeling auralizations are validated by using articulation indexes. As it is well known, one of the main challenges in validating an auralization is how to find an objective metrics to evaluate its quality. The generation of acoustical virtual reality is done by the proprietary computer code RAIOS (Room Acoustics Integrated and Optimized Software), which includes sets of artificial neural networks (ANNs) that models the head-related impulse responses (HRIRs). As output, the computer code furnishes the binaural impulse responses (BIRs) at selected positions in the room. Convolving these BIRs with anechoic signals containing a list of monosyllables, virtual auralizations are provided. These can be applied to subjects to obtain virtual articulation indexes. On the other hand, the articulation test applied in the actual room to the same subjects provides actual articulation indexes. This is done in two university classrooms. The results show that an error lower than 2.5% is found when actual and corresponding virtual articulation indexes are compared. The conclusions are that the auralizations performed by the computer code RAIOS with the ANNs succeed and that the articulation index is a reliable metrics to validate auralizations.