In teleconferencing calls it is difficult to identify the current talker, especially if the person is not well known. In addition, if more than one person is talking at the same time, none of them can be understood easily. In real meetings, however, these limitations are not relevant because humans hear spatially. They take advantage of the "cocktail party effect" to distinguish speakers. In this publication, we assess a teleconferencing solution that we call 3D Telephony, which adds a virtual acoustic room simulation to IP based telephony, thereby achieving a spatial audio experience. We have conducted subjective listening tests of a 3D audio rendering engine to study the impact of the participant's locations in a virtual environment on sound quality, understandability and locatability. Our listening test results show some interesting findings. We identified placements, in which listeners find it very easy to locate virtual talkers and in which their success rate in locating two simultaneous virtual talkers is nearly perfect.
