Vowels, part 4

March 19, 2014

Just So You Know• Today: Source-Filter Theory

• For Friday: vowel transcription!

• Turkish, British English and New Zealand English

• For next Wednesday:

• Production Exercise #3 (on Vowels, natch)

• Formant Measuring Exercise

The Great Lakes Shift• One chain shift is currently taking place in the northern United States.

• Prevalent in Chicago, Detroit, Cleveland, Buffalo, and many places in between

• (but not in Toronto)

• (but maybe in Windsor!)

General Great Lakes

fronting

Hod

200

300

400

500

600

700

800

900

1000

100012001400160018002000

F2

F1

[æ] raising

Had

200

300

400

500

600

700

800

900

1000

10001500200025003000

F2

F1

backing

“ahead”

Hod

600

700

800

900

1000

1100

7009001100130015001700

F2

F1

Female Talkers

Who'ed

300

400

500

600

8001100140017002000

F2

F1

Female Talkers

New Zealand Vowel Shift

http://www.youtube.com/watch?v=JT5AQIlmM0I

A Word of Caution• The vowel system of English can vary greatly from one dialect to another.

• Ex: the vowels of Canadian English have shifted away from their American counterparts…

• (for some, but not all, speakers)

• Shift #1:

• Shift #2:

Unshifted:

Unshifted:

• There are also new shifts underway!

• Shift #3: “head”

• Shift #4: “hid”

• Shift #5: “hood”

Source/Filter Theory: The Source

• Developed by Gunnar Fant (1960)

• For speech, the source of sound = complex waves created by periodic opening and closing of the vocal folds

Source Differences

adult male voice

(F0 = 150 Hz)

child voice

(F0 = 300 Hz)

Just So You Know• Voicing, on its own, would sound like a low-pitched buzz.

• Check out the sawtooth wave spectrum:

• Vowels don’t sound like this because the source wave gets “filtered” by the vocal tract.

“Filters”• For any particular vocal tract configuration, certain frequencies will resonate, while others will be damped.

• analogy: natural variation/environmental selection

• This graph represents how much the vocal tract would resonate for sinewaves at every possible frequency.

Source + Filter = Output

+

=

A Vowel Spectrum

Note:

F0 160 Hz

F1

F2

F3 F4

Output Example: [i]

• Different vowels are characterized by different formant frequencies.

• These reflect changes in the shape of the sound filter.

• (the vocal tract)

Vowel Spectrum #2: [i]

F0 = 185 Hz

F1

F2 F3

at different pitches

100 Hz 120 Hz

150 Hz

Narrow-Band Spectrogram• A “narrow-band spectrogram” clearly shows the harmonics of speech sounds.

• …but the formants are less distinct.

harmonics

Wide-Band Spectrogram• By changing the parameters of the Fourier analysis, we can get a “wide-band spectrogram”

• This shows the formants better than the harmonics.

formants

Wide-Band Spectrogram• By changing the parameters of the Fourier analysis, we can get a “wide-band spectrogram”

• This shows the formants better than the harmonics.

formants

F1

F2

F3

Wide-Band Spectrogram• By changing the parameters of the Fourier analysis, we can get a “wide-band spectrogram”

• This shows the formants better than the harmonics.

formants

F1

F2

F3

voice bars (glottal pulses)

Spectrographically• This is what it looks like when you change the source independently of the filter.

• The formants stay the same, but the F0 and harmonics change.

The Flip Side• This is what it looks like when you change the filter independently of the source.

• The resonating frequencies change, but the F0 and harmonics stay the same.

More Relevantly• In diphthongs, the filter changes while the source can remain at the same F0.

“Boyd”

• Check out the narrow-band spectrogram…

More Music• With (most) musical instruments, we can only change

the frequency of the sound source.

• Timbre is a musical term for the “quality” of a sound.

• I.e., its characteristic resonances.

• E.g., compare the same note played by a trumpet vs. a violin.

• In speech, you can independently change both source and filter frequencies at the same time.

• Like changing the size of a piano…

• As you press different keys on the keyboard.

• This makes the acoustics of speech at least twice as complex as the acoustics of music.

Formant-Reading Tip #1• Another distinction between source and filter characteristics is formant bandwidth.

• Harmonics are exact:

• integer multiples of source frequency

• Resonances are less exact:

• they’re centered around an optimal frequency, but other frequencies may resonate to some extent, too.

• Hence: formants can appear to merge in wide-band spectrograms.

Bandwidth

Bandwidth

Merged Formants

F1

F2

Another Problem: Dynamics

“hod”

F1

F2

• vowel formants are typically not “steady-state” for very long

F1

F2