LIN228H1 Lecture Notes - Velar Nasal, Front Vowel, Spectrogram
75 views11 pages
LIN228H1F 2012 – Week 8 Kochetov-1
Acoustics of English Sounds
This handout discusses only some basic information used in reading spectrograms. Additional example
spectrograms illustrating English sounds can be found in the textbook. See also Chapter 8 of Peter
Ladefoged’s ‘Course in Phonetics’ (http://www.phonetics.ucla.edu/course/chapter8/figure8.html).
A spectrogram is a representation of speech sounds showing time on the horizontal axis and frequency on
the vertical axis. Intensity is shown by the darkness of the representation (darker colouring = greater
Spectrograms of different vowels are distinguished from one another on the basis of formants. Recall that
formants are clusters of harmonics which are enhanced by the resonating properties of the vocal tract. The
configuration of the vocal tract differs for each vowel leading to different formant frequencies for each
vowel. The lowest formant is F1 and F2 give us the most information in distinguishing one vowel from
• F1 is determined by the resonating frequency of the back cavity
o It is lower for high vowels (larger back cavity, lower resonating frequency) and higher for low
vowels (smaller back cavity, higher resonating frequency).
• F2 is determined by the resonating frequency of the front cavity
o It is higher for front vowels (smaller front cavity, higher resonating frequency) and lower for
back vowels (larger front cavity, lower resonating frequency). F2 is even lower for rounded
vowels because the front cavity is increased by lip protrusion.
Back cavity: large
Front cavity: small
Back cavity: small
Front cavity: large
Back cavity: large
Front cavity: large
(and increased by lip protrusion)
(Articulatory speech synthesis: http://www.haskins.yale.edu/; see the last week’s handout for a discussion.)
Peaks in intensity, which are shaped by the configuration of the vocal tract, show up as formants
resonating frequency is higher for a small cavity than for a large one
high front-->low front-->low back-->high back [i-->ae-->a-->u]
F1 starts low and rises until the vowel [a], then F1 falls (inverse relationship to vowel height)
F2 starts high and falls (direct relationship to vowel backness)
F3 is similar to F2, but higher. It depends on F1 and F2
F4 and F5 are not determined by vowel quality
LIN228H1F 2012 – Week 8 Kochetov-2
See Table 8.2 in the textbook for typical English vowel formant values.
Formants show up on spectrograms as dark, relatively horizontal bands.
deed [i] did [ɪ] dead [ɛ] dad [æ]
dude [u] dawd [dɑd]
Diphthongs are vowels that involve a change in vowel quality during their articulation. They can be
recognized on spectrograms due to the change in their formants.
• /aj/ begins with a relatively high F1 on account of the initial low vowel articulation and a relatively
low F2 on account of it being a central vowel. Then the formants move to the low F1 and high F2 of
the high front vowel /i/.
• /aw/ begins similarly as it also begins with a low vowel articulation. The formants then change with
F1 lowering in accordance with the vowel height of the /u/ portion of the diphthong and the F2
lowering as well on account of the /u/ being farther back in articulation than the central /a/.
• /ɔj/ begins with the two formants relatively low and close together and then the spread apart into the
typical _____________ F1 and ____________ F2 of the high front vowel articulation.
buy [baj] bow [baw] boy [bɔj]
F2 is high
F1 is rising
F2 is low
F1 is rising
mid, back --> high, front
LIN228H1F 2012 – Week 8 Kochetov-3
Fricatives are often the easiest segments to pick out of a spectrogram because of the distinctive random
noise pattern in the high frequencies.
• The labiodental and dental fricatives are generally weak with a much lower intensity than the
sibilants. The sibilants /s, z, ʃ, ʒ/ have greater intensity.
• The alveolars /s/ and /z/ have random noise pattern visible in the 4000 to 8000 Hz range with the
postalveolars having somewhat lower frequencies in the 2000-6000 Hz range.
/h/ has been described in this course both as a voiceless fricative and as a voiceless vowel. It has weak
formant patterns that we associate with the resonant properties of vowels and it has weak random noise
patterns as fricatives do. It shows up on spectrograms as a noise pattern in the formant frequencies of the
See the spectrograms below for examples of [f] and [h]. See last week’s handout (Appendix) for
spectrograms of sibilant fricatives.
fur [fɚ] hay [hej]
During stops no air passes through the vocal tract as there is complete closure. Stops thus show up as a gap
on a spectrogram with no apparent sound. Clues to the place of articulation of stops are found in the
transitions to the surrounding vowels.
• Labial consonants lower F2 of surrounding vowels.
• Alveolar consonants have relatively level formants in the vowel transitions (with some variation
depending on vowel backness/frontness).
• Velar consonants bring raise F2 and lower F3 bringing F2 and F3 together.
Voiceless stops can be distinguished from voiced stops due to the presence of a low frequency voice bar in
the voiced stops. This can, however, be very faint. Aspirated voiceless stops end in a burst of high-
points up into the gap
points down into the gap
voiced fricatives often show vertical striations, and sometimes a voice bar (formant-like, low frequency bands)
(aperiodic wave form)
The release of a stop is shown by a vertical spike (high energy noise)