Novo P.

Speech produced by crowds is often the dominant sound source in places such as bars and restaurants and therefore has a major impact on the aural comfort perceived by the crowd. Although room shape and materials have a major influence on the final sound field, the set of distributed sound sources that constitute the speakers in the crowd has a major impact. One aspect, is that they collectively form a more diffuse sound field than would have been generated by a single sound source in that same room. Recent work in the area of hearing in diffuse sound fields (DSFs) has revealed that under specific conditions the auditory events locations do not equally surround the listener although the sound field does. The frequency range where the frequency-dependent cross-correlation of the left and right ear signals of a listener immersed in a DSF assumes negative values has been proposed as contributing to explain this phenomena. In this paper the question of preference in DSFs generated by crowd speech is addressed. The study involves presenting the listeners with binaural signals that were recorded under laboratorial and real world conditions as well as the presentation of synthesized sound fields. The parameters studied include DSF dimensionality (2D and 3D) and non- perfectly DSFs. The main focus of the analysis is on the relation between the cross-correlation of the signals present at each ear, in particular in the frequency range where it assumes negative values, and preference.