Comments on, "Active Fluid Dynamic Voice Production Models, or There is a Unicorn in the Garden," Herbert M. Teager and Shushan M. Teager.
--------------------------------------------------------------------------------------------
Above is the talk given by Herb Teager to explain the necessity for new model for speech production (focus on non-linear aspect of speech flow).
Herb Teager gives evidence that the speech flow is not at all planar and that it is "separated flow."
Following things bothered Teager the most about the way modeling of speech has been done,
1. About less than 1% of the equivalent mechanical lung input energy is involved in the rate of change of volume velocity at the glottis. The 99.5% of the energy is "used up" somewhere, but the research community is NOT bothered about it.
2. Representing the speech apparatus as a "passive linear system" is not the right one, as the time domain and frequency domain observations aren't completely interchangeable.
3. Related to the behaviour of speech signal under different media (the famous helium effect), the relative shift of the formants (usually seen for the formants below 200Hz) should be proportional to the change in the density, and to an extend related to the atmospheric pressure. But, Teager observed the shift / changes in the speech signal in the other direction, and by a different factor (by the factor of square root of the density).
4. The data (and hence the observations made) obtained using the hot-wire anemometry indicate, (a) "uniform, plane acoustical wave were incompatible with the observed separated flows", (b) flow is separated, possess rotations and doesn't necessarily repeat themselves across the cycle, (c) flow patterns remain consistent for a given vowel, but are radically differ across the phonemes.
5. Flow pulses arising from the front and back of the mouth move at different speeds and attenuations. Most importantly, the pressures were uniform over cross sections within which separate flow occurred. The driving mechanism for the separate flow is still unknown.
6. Acoustic impedance which relates pressures and flows in the sound wave does not translate the flow wave that moved without a corresponding pressure wave.
*** Teager assumes that soliton (and "momentum wave" a termed coined by him) can help explain the phenomenon.
7. The continuity equation and motion equations used to model the speech flow are one dimensional (or can be extended to two-dimensions) but cannot be applied to 3-D and most importantly separated flow, as the equation is over-simplified version. "The basic unsimplified Navier-Stokes equations are intrinsically unstable and unsolvable."
The flow should be considered as "separated flow" made up of threads and gusts which change quickly in direction, time and space, i.e, asymmetry of wave propagation in separated flow.
Then Teager went on the discuss about the Models that he thinks will help us model all or some of the above anamolies.
The basic premise is to omit the effects in the lungs, and just concentrate on the effects or air-flow interactions at the glottis and in the mouth.
The major focus will be to understand the non-linear energy feedback and modulations effects that occur in a jet-cavity interactions, dynamics of rotating vortex flows.
1. A flow leaving a constriction, like a JET, with the cavity (mouth/glottis) being asymmetric, the jet will differentially exhaust one or the other side and willl be thus "attracted" to the nearest wall, but the attachment may be unstable or oscillatory. This separated flow will simulataneously give rise to vortexes (axial, radial or a combination of both).
2. The the complex system at glottis, as glottis is providing a time-varying, directional source of separated air flow in the presence of vortexes modifying those flows with their own internal and external synamics.
3. A jet-cavity interaction can result in:
a) Oscillator when there is regenerative addition of energy when the energy is already high or removal when it is low. Rotating flow can store kinetic energy, and then return it as pressure, with no differential flow. Also the interaction with the separated flows acts as a non-linear yielding wall. In almost all the cases, because of the radial, rotating vortex within, the oscillator exhibits a strong amplitude modulation at the internal vortex precession rate.
b) the axial vortex at the converging outlet - acts as non-linear plug. When pressure in the chamber is high, outward flow is impeded by the compressed outlet vortex, but when pressure is low, the vortex expands and allowsd relatively more exit flow and an increase in vortex strength. Causing oscillations.
If two flows modulated with by different frequencies collide, (with the collision taking place over part of a pitch period), then all sorts of combinations of frequencies will result.
c) if for some reason, the regenerative coupling is small or happens over a part of cycle, then the oscillator will degenerate as a filter.
Teager suggested:
1) to have or perform more experimentation on humans to understand the whole gammit or jet cavity interaction, or similar concept, within the different areas of mouth, throat and the whole system coupling.
2) avoid the computer simulations of the "simplified Navier-Stokes equations".
Teager during the discussion that followed suggested that:
Mask-type pneumotachograph needs to calibrated for pulsatile flows (as these flows occur during speech production) to access the flow variations with instruments other than hot-wire anemometer.
The nominal range for measurements or flows within the mouth / throat:
Frequency range: 0 -4 KHz
Physiological range: 10-300 cm/sec
Frequency response of the instrumentation usd by Teager is float (in 0 - 5KHz) and is independent of the flow rate.
-----------------------------------------------------------------------------------------------
*** we have affirmed that the seven dimensional (three linear velocity and three angular velocity, plus time) flow patterns are unique for each vowel, and have published four sets of manifestly different trajectory data to back up this postulate.
-----------------------------------------------------------------------------------------------
No comments:
Post a Comment