-
SOURCE FILTER THEORY CHARACTERISTICS
- from vf vibratin
- harmonic spectrum
- -energy at harmonic intervals
- -usually some noise in btw, though for normal voices there is very little noise at lower frequencies (below 3-4 khz)
- based on the configuration of the upper vocal tract
- modifies energy from the source
- two parts
- -transfer function (ie formants)
- -radiation characteristic from lip opening - +6 dB/octave
-
harmonic spectra and women
- female harmonics are spread farther apart than males
 femalemale
-
what is the SOURCE in the source filter theory
vocal folds (source function)
-
what is the FILTER in the source filter theory
- vocal tract
- allow some resonance to go through while some is restricted (shown as transfer function, peak=resonate, trough=restricted
-
what is the OUTPUT in the source filter theory
- lips
- combo of source function and transfer function
-
what are source, filter and radiation in relation to one another?
- source and radiation characteristics are relatively independent of articulatory configuration
- *radiation refers to lips
-
what is transfer function highly dependent on
articulatory configuration
-
formant frequencies
do not uniquely specify all vowels for perception because there are individual differences, not the same for all (ie how one says /a/ may be dif for another person but still recognized as /a/)
-
what are the characteristics of front vowels?
- large separation of f1 and f2
- small separation of f2 and f3
- quantal vowel /i/ - high, front vowel (spread lips)
- -lowest f1
- -highest f2
- -highest f3
-
what are the characteristics of back vowels
- small separation btw f1 and f2
- large separation btw f2 and f3
- quantal vowel - /u/ - high back vowel (rounded)
- -low f1
- -lowest f2
- -lowest f3
-
characteristics of central vowels
- evenly spaced
- mid range value of F1
-
where is the optimal place to measure formant frequency
in the center of the vowel to obtain steady state info
-
LPC
Linear predictive coding gives envelope of the formants (frequency by amplitude)
-
spectogram
depicts all speech sounds as energy. shows freq, amp and time
-
wide band spectogram
- identifying formants and noise
- generally used for speech analysis
-
narrow band spectograms
identifying harmonics
-
how can formant frequencies be displayed
-
what does lip rounding do?
- ellongates the length of the tube so lowers the formant freqs
- usually back vowels have lip rounding
- front vowels usually do not
-
how do you measure amp and bandwidth
- can use LPC analysis
- amp is measured at the peak of the formant
-
what is the bandwidth
- freq range of energy of a sound
- depends on the damping within the vocal tract
-
what are nasalized vowels
- formants from oral cavity
- formants from nasal cavity
- anitformants due to bifurcation
- damping due to soft, mucous tissue in nasal passages
- lower amp than non nasalized vowels
-
tense vowels
press articulators tightly and hold position for a long time
-
lax vowels
assume a looser constriction and held for short times
-
how does speech rate affect formants
- articulatory undershoot occurs more as speech rate increases
- some studies show a change in FF some have not
- duration of vowels is reduced with increased speech rate
-
consonant formants
•Smaller vocal tract constriction than vowels
–Vary by degree of constriction
•Classified by manner, place, and voicing
–Manner: Glide, liquid, nasal, stop, fricative, affricate
–Place: Bilabial, Alveolar, Velar, etc.
–Voicing: Voiced or voiceless
-
what consonants are similar to vowels u and i
–Duration of constriction typically shorter than for corresponding vowels
–May be more extreme, constriction tighter than for vowels
•Ex: /w/ oral constriction is tighter than for /u/
•Ex: oral constriction is tighter for /j/ than for /i/
–Glide into and out of constriction
|
|
