Contribution
Towards Contemporary Edison Tests - Part 2: Assessing the plausibility of speech reproduction in anechoic conditions
* Presenting author
Abstract:
Plausibility has emerged as a key metric for evaluating audio in augmented reality (AR) environments. While recent studies have extensively examined the plausibility of binaural synthesis for loudspeaker-based sound sources, it remains unclear how these findings apply to non-loudspeaker sources, such as the human voice. The human voice is an important sound in our everyday lives and hence essential for AR applications. This study presents preliminary results from two listening experiments conducted in anechoic conditions, focusing on the plausibility of speech reproduction. In both experiments, participants were asked to identify whether a speech sample was produced by a human talker behind a curtain or virtually reproduced. In the first experiment, the virtual reproduction was a loudspeaker playing back anechoic speech recordings. In the second experiment, the virtual reproduction was achieved through convolution with head-related transfer functions (HRTFs). The experiments are a starting point for further research on the plausibility of human speech in AR scenarios.