Singing in Space(s): Singing performance in real and virtual acoustic environments - Singers' evaluation, performance analysis and listeners' perception

The Virtual Singing Studio (VSS), a loudspeaker-based room acoustic simulation, was developed in order to facilitate investigations into the correlations and interactions between room acoustic characteristics and vocal performance parameters. To this end, the VSS provides a virtual performance space with interactivity in real-time for an active sound source - meaning that singers can hear themselves sing as if in a real performance space.

An objective evaluation of the simulation was carried out through measurement and comparison of room acoustic parameters of the simulation and the real performance space. Furthermore a subjective evaluation involved a number of professional singers who sang in the virtual and real performance spaces and reported their impressions of the experience. Singing performances recorded in the real and virtual spaces were compared via the analysis of tempo, vibrato rate, vibrato extent and measures of intonation accuracy and precision.

A stimuli sorting task evaluated listeners' perception of the similarity between singing performances recorded in the real and simulated spaces. A multi-dimensional scaling analysis was undertaken on the data obtained and dimensions of the common perceptual space were identified using property fitting techniques in order to assess the relationship between performance attributes and the perceived similarities. In general significant proportions of the perceived similarity between recordings could be explained by differences in global tempo, vibrato extent and intonation precision. Although there were few statistically significant effects of room acoustic condition all singers self-reported changes to their singing according to the different room acoustic configurations, and listeners perceived these differences, especially in vibrato extent and global tempo.

The present VSS has been shown to be not fully ``realistic'' enough to elicit variations in singing performance according to room acoustic conditions. Therefore, further improvements are suggested including the incorporation of visual aspect to the simulation. Nonetheless, the VSS is already able to provide a ``plausible'' interactive room acoustic simulation for singers to hear themselves in real-time as if in a real performance venue.