A speech corpus for multitalker communications research.

Several recent experiments at the Air Force Research Laboratory have investigated the utility of spatial audio displays for augmenting speech intelligibility in multitalker communications environments ~Bolia et al., 1999; Nelson et al., 1998a; Nelson et al., 1998b; Simpson et al., 1999!. Some of the goals of this research included: ~1! an empirical determination of the maximal number of channels for which the benefits of spatialization may be realized in a spatial audio display designed to aid in the segregation of simultaneous, context-independent speech sources; ~2! an evaluation of the efficacy of four different spatialization schemes for this task; and ~3! the manner in which these factors interact with the sex of the target talker. In order to accomplish these goals, a large number of speech samples from talkers of both sexes were required. The purpose of this article is to describe the methods employed in collecting these speech samples, as well as the form of the resulting corpus, with the intent that other researchers in the field might benefit from their availability.