Abstract
This paper presents a perceptual experiment aimed at assessing the spatial quality of acoustic environ- ment rendering using a 4th order ambisonic auralization system. A novel test protocol is developed for this purpose, based on comparing the perceived spatial attributes of sound sources in both real (in-situ) and virtual listening conditions (loudspeaker-based ambisonic auralization of measured SRIRs). The perceptual evaluation is conducted using a specific reporting method combined with a virtual reality interface, enabling simultaneous assessment of perceived distance, angular position, and apparent width of sound sources. The test is conducted in three "office like" rooms, varying in reverberation properties and size. The results highlight differences in spatial perception between a) real rooms and b) their reproduction through the au- ralization system. Overall, localization performance is worse in auralized conditions than in real conditions, as evidenced by a clear increase in localization errors in azimuth and elevation, along with an increase in reported source width. This study also reveals that the spatial accuracy of the auralization depends on the rooms being auralized.
Sound stimuli
For the auralization condition, stimuli are obtained by convolving the three anechoic sound excerpts (Speech, Guitar, Burst) with the measured Spatial Room Impulse Response (SRIR) for each room (S1, S2, S3) and each source position (D2, D4). The stimuli presented in the table below are binaural conversions of the original 4th-order ambisonic stimuli.
 |
 |
 |
|
|---|---|---|---|
| Room 1 (S1) | Room 2 (S2) | Room 3 (S3) | |
| 2m (D2) | |||
| 4m (D4) |