Abstract
Speech directivity induces variations of the amplitude and spectrum of the radiated sound with the direction. It is gaining interest for the rendering of speech in three dimensional environments (real or virtual), but it is also related to more fundamental research questions, such as the intelligibility with competing speech (cocktail party problem). Speech directivity is most often quantified by averaging in octave bands the speech production of real subjects recorded simultaneously at different locations with microphone arrays in anechoic environments. Due to the variability of the physical mechanisms of speech production, the radiation patterns differ between different speech sounds. However, a part of the observed variability may be due to the averaging process itself, which is influenced by the spectral differences between the different speech sounds. In order to investigate to what extent and in which frequency range this variability is actually due to differences in directionality, directivity patterns are computed in narrower frequency bands with constant width. The details revealed by this higher frequency resolution also allow one to identify the expected influence of the dimensions of the subjects, the mouth opening and the contribution of the nasal cavity to the sound radiation. Normalized octave band averages are computed and compared with the commonly performed octave band averages.
Original language | English (US) |
---|---|
Title of host publication | Forum Acusticum |
Subtitle of host publication | Proceedings of the FA2020 Conference |
Place of Publication | Lyon, France |
Pages | 2027-2033 |
Number of pages | 7 |
DOIs | |
State | Published - Dec 2020 |
Event | Forum Acusticum 2020 - Lyon, France Duration: Apr 20 2020 → Apr 24 2020 |
Conference
Conference | Forum Acusticum 2020 |
---|---|
Country/Territory | France |
City | Lyon |
Period | 4/20/20 → 4/24/20 |
Keywords
- Speech
- Directivity