VOWELS attempts to impose the formant characteristics associated with vowels on to the sound, in a time-varying manner. PARAMETERS vowels Datafile of times (in secs., increasing from 0) and vowels, the latter coded as: ee: as in
'heat' i: as in
'hit' e:
as in
'pet' ai: as in 'hate'
half width
of formant
peaks in Hz, as fraction of formant centre frequencya: as in 'pat' ar: as in 'heart' o: as in 'hot' or: as in 'taught' oa: as in 'boat' u: as in 'hood' oo: as in 'mood' xx: as in Southern English 'hub' x: neutral vowel in 'herb' or 'the' (The implied pronunciations here are English.) Example: 0 ai 0.7 ai 0.72 ee 1.44 o 2.17 oa The vowel data values are continuous from one time to the next; e.g. the vowel from time 0 to 0.7 is static, but between 0.72 and 1.44 changes from ee to o. Range:
0.01-1-10
Default: 0.25
steepness Resonance
of
formant peaks, similar to 'Q' in filtering (smaller is
narrower; more
sharply defined). Range: 0.01-1-10 Default: 3.0
peak range Ratio
of
(max) peak height to (max) total range of peak
height Range: 0-1,
default 0.95.
threshold Proportion
of
spectral window level compared with vowel envelope level
Range: 0-1,
default 0.5.
NOTES To be convincing, the source sound must have a wide range of frequencies. For example, it is not possible to bring out the vowel 'oo' (formant centres 320 and 800 Hz), if there are no frequencies present below 1000Hz. Most of the parameters, apart from vowels, are over-detailed and work well enough with their default values. Steepness may be worth setting to control the width of the peaks. The program clearly succeeds in colouring the sound, but direct vocoding with a spoken/sung vowels may often be more convincing. Also note the sister program Pitch-Vowels.
|