Artword & Speaker: To Sound...

A command to synthesize a Sound object from the selected Speaker and the selected Artword.

This is the command that performs the actual articulatory synthesis. See Articulatory synthesis.

Settings

Sampling frequency (Hz)
the number of times per second that the equilibrium widths and lengths and the tensions of the muscles are recomputed from the Artword. This will also be the sampling frequency of the resulting sound and of the optional resulting tube widths, air pressures, and air velocities. The standard value is 22050 Hz.
Oversampling
the number of times that the aerodynamic quantities and the state of the tube walls will be recomputed during each sample period. The standard value is 25.
Width 1, Width 2, Width 3
the numbers (see below) of the tubes whose widths you want to monitor. E.g., if Width 1 is 36, the synthesizer will create a Sound object named width36, which contains the width of tube 36 (the lower glottis) as a function of time, expressed in metres. To prevent the creation of a width object, specify 0 (the standard value).
Pressure 1, Pressure 2, Pressure 3
the numbers (see below) of the tubes whose air pressures you want to monitor. E.g., if Pressure 3 is 37, the synthesizer will create a Sound object named pressure37, which contains the air pressure of tube 37 (the upper glottis) as a function of time, expressed in Pascal. To prevent the creation of a pressure object, specify 0 (the standard value).
Velocity 1, Velocity 2, Velocity 3
the numbers (see below) of the tubes whose air velocities you want to monitor. E.g., if Velocity 1 is 60, the synthesizer will create a Sound object named velocity60, which contains the air velocity of tube 60 (in the mouth) as a function of time, expressed in metres per second. To prevent the creation of a velocity object, specify 0 (the standard value).

Stability

The internal sampling frequency for the aerodynamics is the specified Sampling frequency, multiplied by the specified Oversampling. With the standard settings, this is 22050 times 25 = 550750 Hz.

To ensure the stability of the synthesis, this internal sampling frequency should not be less than the velocity of sound (353 m/s) divided by the length of the shortest tube. For the standard Female, Male, and Child speakers, the shortest tube is the upper glottis, which has a length of 0.7, 1.0, and 0.3 millimetres, respectively. The minimum internal sampling frequencies, therefore, are 504286, 353000, and 1176667 hertz, respectively.

Time resolution

To capture the microscopic pressure changes in the glottis, you will want maximum time resolution. For a female speaker, you could set Sampling frequency to 550750 Hz, and Oversampling to 1.

Tube numbers

Here are the tube numbers that you can use for the Width, Pressure, and Velocity settings:

1..23: lungs (from bottom to top)
24..29: bronchi (from bottom to top)
30..35: trachea (from bottom to top)
36: lower glottis
37: upper glottis (not for a one-mass model)
38..49: pharynx (from bottom to top)
50..51: nasopharyngeal branching
52..64: mouth (from back to front)
65..78: nose (from back to front)
79..86: conus elasticus (only for a 10-mass model)
87..89: glottal shunt between the arytenoids (from bottom to top)

Some structural properties:

• Tube 1 is closed at the bottom.
• Tubes 64 (lips) and 78 (nostrils) radiate into the air.
• The nasopharyngeal branch is at tubes 50, 51, and 65. They are constrained to have equal lengths.
• For a one-mass model of the vocal cords, tube 36 is connected to 38.
• For a 10-mass model, tubes 32..35 are replaced with 79..86, so that tube 31 is connected to 79, and 86 is connected to 36.
• A glottal shunt will be implemented if the speaker’s shunt.Dx attribute is not zero. A branch is then made from tubes 34 and 35 (or 85 and 86) to 87, and from tube 89 to 38 and 39.

Links to this page


© ppgb 20040331