The sound localization ability of humans is dominantly based on temporal differences and sound-pressure-level differences between the ears. Localization is most accurate with a single sound source in the frontal horizontal plane. However, natural scenarios typically consist of more than one sound source having narrower or broader spatial extent. The aim of this thesis was to study how humans perceive such complex scenarios and how various attributes of the sound scene contribute to the perception. In all of the listening experiments of this thesis, the perception of widely distributed sound scenarios was studied on the horizontal plane. Sounds were presented from varying loudspeaker setups with up to thirteen loudspeakers close to one another. With noise signals, the directions of only up to three individual sound sources were accurately perceived at their original locations, and widely distributed sound source groups were perceived as narrower than they were. With monophonic music and ambient-noise signals that were synthesized to be spatially extended, the perception was found to be yet slightly narrower than with the noise cases. The effects of changes in the temporal domain were investigated as well. First, peaks in the signal envelope were shown to aid in the directional discrimination of a sound in the presence of two other simultaneous spatially non-overlapping sounds. Second, accurate spatial perception of temporally successive sounds was shown to be difficult, as the inter-stimulus-interval was required to be high in order to perceive the correct spatial distribution of the presented sounds. In addition to studying human perception, a binaural auditory model was developed to mimic human spatial hearing performance. The design of the model was motivated by knowledge on neurophysiology and psychoacoustics. A binaural activity map produced by the model showed performance matching that of humans in various listening scenarios. Importantly, spatially wide or complex sound scenarios can be analyzed as well, a feature uncommon in auditory modeling. Furthermore, the model was applied in the assessment of spatial sound reproduction techniques and showed differences in activation when there were perceivable differences in the reproduced scenarios. The results give new insight on spatial sound perception. Overall, spatial hearing is not accurate in localizing multiple simultaneous sounds as opposed to accuracy in vision. Obtained results help in the development of parametric spatial sound reproduction techniques, as the techniques do not need to reproduce details that are not perceived by human listeners.
|Translated title of the contribution||Tilajakaumaltaan monimutkaisten äänitapahtumien havaitseminen ja havaitsemisen mallintaminen|
|Publication status||Published - 2015|
|MoE publication type||G5 Doctoral dissertation (article)|
- sound localization
- spatial hearing
- auditory modeling