New considerations for vowel nasalization based on separate mouth-nose recording

In this paper an experimental setting is described which allows to separately record the speech signals emitted from the mouth and the nostrils. A series of transitions from an oral configuration to the nasopharyngeal configuration is then recorded with a speaker capable to control the movement of his velum while minimizing that of other articulators. The analysis of the recorded signals in the light of simulations with an acoustic model clearly shows that the formant evolution pattern observed in the mouth output is essentially caused by a modification of the oral tract shape due to the velum lowering and the pharyngeal constriction, while the connection with the nasal tract can be neglected in first approximation. On the other hand, the velum movement controls rather the nose output amplitude than its spectral pattern.