* -*- Mode: outline -*-
* Vocal tract resonances
organ-pipe vocal tract model
Tube with sound source at one end.
Resonance at wavelengths 4L, 4L/3, 4L/5, ...
Correspond to frequencies c/4L, 3c/4L, 5c/4L, ...
L is the length of the tube.
c is the speed of sound in air. c=340 m/s
Typical tract 17 cm.
Resonances at 500 Hz, 1500 Hz, 2500 Hz, ...
Vocal tract resonances produce peaks, called formants.
Formant one, the lowest, 200 Hz to 1000 Hs.
Formant two, 500 Hz to 2500 Hz.
Formant three, 1500 Hz to 3500 Hz.
* Sound Sources
** Voiced
Air passing by the vocal cords causes them to flap
ah, uh, etc.
** Asperation
Folds of the larynix held slightly apart,
to make air passing through between them turbulant. 
h
Not normally voiced
** Frication
Turbulance between tongue and top of mouth
ss, sh, f
May be voiced
** Source filter model
* Classification of Speech sounds
** Phonemic transcription
Phonemes are logical, language relative units.
Semantic role as word discriminators.
** Categories of speech sounds
*** Vowl
uh a e i o u aa ee er uu ar aw
*** Dipthong
[not classified as individual phonemes]
*** Glide (or liquid)
r w l y
*** Stop
correspond with nasals
**** Unvoiced
p t k
**** Voiced
b d g
*** Nasal
correspond with stops
m n ng
*** Frictive
**** Unvoiced
ch
**** Voiced
j
*** Aspirate
h
** Coarticulation
A sound is effected by those that come on either side of it.
** Prosody
Global attributes of speech.
Prosodic, or suprasegmental features.
*** Features of voice quality
paralinguistic
anatomical differences
*** Features of voice dynamics
**** pitch or fundamental frequency
pattern of pitch variation is intonation
**** time
rhythm of speech and the overall tempo
**** amplitude
works with rythm to produce stress
** Speech storage
*** Sampling
Discretation in time.
Sampling interval of T seconds.
Sampling frequency is 1/T Hz.
**** Aliasing
Componants 1/2T+f, 3/2T-f, 3/2T+f, ... masquerade as a component at 1/2T-f.
Must filter the frequencies greater than half the sampling frequency
to avoid aliasing.
Phone system samples at 6.8 kHz.
*** Quantization
Discretation in amplitude.
*** Logarithmic quantizatino
Better signal to noise ratio over a wider range of input amplitudes.
**** Companding
Compressing-expanding