Acoustic Tube Endoscopy – Virtual tube model
In addition to identifying tube openings by evaluating the changes in transmitted energy, a virtual time variant tube model can be calculated from the impulse responses gained in the system identification. For this we exploit a tube model based on the one known in speech processing, where it is used to model the speech production system. The calculation of the tube parameters is performed mainly by determining the autocorrelation function of the impulse response and the reflection coefficients at the junctions of the tube segments by means of the Levinson-Durbin algorithm.
But there are significant differences between the vocal tract and the transmission link between nose and ear, that must be taken into account. One major aspect is the acoustical impedance at the end of the acoustical path. For the vocal tract we assume, that the mouth is open when we speak thus the impedance is zero. Looking at the transmission path from ear to nose and considering that the miniature microhpne placed in the outer ear canal seals this side of the transmission path, the impedance here is very high. The parameters of the tube model have to be adjusted appropriately.
The figure below shows the calculated tube at a time instance at which the eustachian tube was open. The red line represents the significant sections of the real anatomy. As you can see there is a significant correlation between the two.
