This page discusses glides and liquids, which together make up the group of sounds known to phoneticists as approximants. Various sources point out that while sounds in this group are produced like vowels, they function mostly like consonants. For more background on approximants, see the second half of James Kirby's PowerPoint, “Nasals, approximants, rhotics, and laterals”.
was in we (labial) and
jas in you (palatal). The two glides are effectively diphthongs with faster transitions (100 msec or shorter as compared to 300 msec for diphthongs). According to Felicity Cox, “If you extend the transition period, the semivowels
jwill sound more like the vowels
idiphthongising with the following vowel.”
l(lateral). In contrast to the glides, liquid sounds rely heavily on the third vocal formant F3 to distinguish them from regular vowels.
||labial glide||Klatt||290||610||2150||The source amplitude … should be about 10 dB less than in the vowel.|
|Cox||250-450||600-850||2000-2400||Very short steady state component. 30 msec is sufficient.|
||palatal glide||Klatt||260||2070||3020||The source amplitude … should be about 10 dB less than in the vowel.|
|Cox||huge, opulent, senior; German ja||200-300||1850-2100||2620-3050|
||rhotic liquid||Klatt||310||1060||1380||The source amplitude … should be about 10 dB less than in the vowel.|
|Cox||screw||300-350||1000-1200||1600-1750||ɹ has a steady state component of at least 50 msec up to about 300 msec possible|
||lateral approximant||Klatt||310||1050||2880||The source amplitude … should be about 10 dB less than in the vowel.|
Table 3 summarizes the fruits of my internet search for synthesis parameters relating to glides and liquids.
Some entries are less convincing that others; in particular, Klatt's
l sounded more like an
me. Cox prefers to list ranges rather than specific values, which while academically more correct is frustrating for my
Klatt's point about scaling back the amplitude for liquids makes sense when the sounds are used like consonants (as in already and around) but less sense when the sounds are used like vowels (as in all or for). Also, I found his recommended reduction of 10 dB distruptive. In Listing 5 the amplitude used for glides and for liquids-used-like-consonants is 1900, which is only 5 dB down from the amplitude of 6000 used for vowels.
Listing 5 presents the third-iteration synthesis of “Daisy Bell”.
The notes are exactly the same as in Listing 2
and Listing 4. However, the single “Amplitude”
ramp that initiated both previous
listings is replaced, now and going forward, by many such ramps, dispersed throughout the text.
Where words employ glides and liquids, new ramps controlling “Formant 1”, “Formant 2”, and “Formant 3” have been interposed. These new indications for glides and liquids are color-coded in blue.
The character of a glide is essentially transitional. This is unlike a vowel, which is static, and more like a diphthong.
However glides differ from diphthongs in that the transition takes place over a much shorter time interval. Where a diphthong
might take 300 msec., a glide might take half that time or less. Consider for example the opening
“your”. I have coded this using two segments:
j. This segment starts at time 14.00 and holds for 50 msec.
jto the formants listed in Table 1 for the neutral vowel
ə. This segment starts at time 14.05 and unfolds for 150 msec.
Between these two segments it's the transition which is most dwelled upon. Which is reenforced by the fact
that much of the steady-statement is taken up by
note #19's 30 msec attack duration, during which the amplitude
ramps up linearly from zero.
Liquids, by contrast, happen by tongue contortions which push formants into ranges not visited by vowels.
If the liquid is positioned at the beginning of a syllable (for example, the second syllable of “marriage”
at time 55.00), then the steady-state and the out-of transition have equal durations.
However when a liquid is positioned at the end of a syllable (for example the final
“your”), then a speaker would tend to dwell on the steady-state phase.
For the final
r in “your” I have coded a 150 transition from
ɔɹ and a 400 msec. steady-state (although the segment lasts 500 msec.,
note #19 releases 100 msec. before the end of the segment).
For liquids, nasals, and plosives the formant transition moving into the consonant is sometimes referred to as the “onglide”, while the formant transition out of the consonant is sometimes referred to as the “offglide”. I like these two terms in spite of whatever confusion they may cause, so will adopt them going forward.
Next topic: Nasals
|© Charles Ames||Page created: 2014-02-20||Last updated: 2017-06-12|