Recap In the first lecture we distinguished periodic sounds - - PowerPoint PPT Presentation

▶
recap
SMART_READER_LITE
LIVE PREVIEW

Recap In the first lecture we distinguished periodic sounds - - PowerPoint PPT Presentation

Dec 05, 2023 471 likes •625 views

Recap In the first lecture we distinguished periodic sounds Spectral features of fricatives and stops (same patterns repeats every cycle) from aperiodic sounds (no regularity in the pattern of air perturbations). C s m v hc v

slide-1
SLIDE 1

Spectral features of fricatives and stops

Cơ sở âm vị học và ngữ âm học Lecture 14

Recap

  • In the first lecture we distinguished periodic sounds

(same patterns repeats every cycle) from aperiodic sounds (no regularity in the pattern of air perturbations).

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55

Fricatives

  • Sound source: aperiodic noise →

random pattern in the waveform. (không theo mô hình)

Amplitude

d ʒ ɔ n s n̩

Time (s)

Fricatives

  • In a spectrogram, fricatives feature energy over a wide

frequency range.

6000 Frequency (Hz) Time (s) 0.47

d ʒ ɔ n s n̩

slide-2
SLIDE 2

Voiceless fricatives

  • The further back the constriction, the longer the front

cavity, and the lower the peak in spectral energy.

From Ladefoged (2006), A Course in Phonetics

Voiceless fricatives

  • For [f T] there is less filter → weaker energy

From Ladefoged (2006), A Course in Phonetics

Voiceless fricatives: overview

  • [s] and [S] have stronger energy because they are

sibilants (âm xuýt) – the air encounters a sharp obstacle (e.g., edge of teeth) Location of energy maximum độ năng lượng [f] between 3 and 4 kHz weak [T] above 8 kHz weak [s] between 5 and 10 kHz strong [S] between 3 and 5 kHz strong

Voiceless fricatives: overview

  • For [f T] the constriction is in the front of the oral cavity,

so there is hardly any vocal tract in front of the constriction to filter the sound! Location of energy maximum Energy strength [f] between 3 and 4 kHz weak [T] above 8 kHz weak [s] between 5 and 10 kHz strong [S] between 3 and 5 kHz strong

slide-3
SLIDE 3

Voiced fricatives

  • [v D z Z] are spectrally similar to [f T s S], but with

striations/voicing bar and weaker frication.

From Ladefoged (2006), A Course in Phonetics

Voiced fricatives

  • Voiced fricatives (x-linguistically rare) are difficult to

produce, and are frequently voiceless or ‘partially voiced’

From Ladefoged (2006), A Course in Phonetics

Other clues

  • What (else) distinguishes these fricatives?

From Ladefoged (2006), A Course in Phonetics

[h]

  • Listed as a fricative in the IPA chart, but actually...
  • The source of sound is not air being forced through a

narrow opening, but instead caused by air rushing through the open vocal folds and vocal tract

  • This is the same sound source as whispered speech
slide-4
SLIDE 4

[h]

  • Listed as a fricative in the IPA chart, but actually...
  • The source of sound is not air being forced through a

narrow opening, but instead caused by air rushing through the open vocal folds and vocal tract

  • This is the same sound source as whispered speech

[h]

  • Listed as a fricative in the IPA chart, but actually...
  • The source of sound is not air being forced through a

narrow opening, but instead caused by air rushing through the open vocal folds and vocal tract

  • This is the same sound source as whispered speech

he ha! who

From http://home.cc.umanitoba.ca/$\sim$robh/howto.html#intro

‘Stole my house’

slide-5
SLIDE 5

Stops

  • Stops consist of two parts: a closure and a noise burst.

Time (s) 0.738 5000

Stops

  • Stops consist of two parts: a closure and a noise burst.

Time (s) 0.738 5000

Stops

  • Stops consist of two parts: a closure and a noise burst.

Time (s) 0.738 5000

Voiced stops

  • In voiced stops, we see a ‘voice bar’ during the closure
  • these are weak striations at the low end of the frequency

scale (because the mouth is still closed).

Time (s) 0.7848 5000

slide-6
SLIDE 6

Voiced stops

  • In voiced stops, we see a ‘voice bar’ during the closure
  • these are weak striations at the low end of the frequency

scale (because the mouth is still closed).

Time (s) 0.7848 5000

Voiced stops [b d g]

  • The burst of voiced stops often does not appear clearly

in a spectrogram

  • If closure and burst are similar for all voiced stops, how

do we distinguish e.g. [b] from [d]?

Voiced stops [b d g]

  • The burst of voiced stops often does not appear clearly

in a spectrogram

  • If closure and burst are similar for all voiced stops, how

do we distinguish e.g. [b] from [d]?

Formant transitions

  • At the moment the stop constriction is released, the

resonances of the vocal tract filter change very rapidly.

  • These changes are called formant transitions.
  • Formant transitions are extremely important perceptual

cues to phonetic contrasts.

slide-7
SLIDE 7

Formant transitions

  • In fact, work from speech synthesis shows you can’t

understand speech without the formant transitions – even though they only average about 50ms in duration.

  • What does this mean for the conception of words as

strings of phone[mes]?

Formant transitions

  • In fact, work from speech synthesis shows you can’t

understand speech without the formant transitions – even though they only average about 50ms in duration.

  • What does this mean for the conception of words as

strings of phone[mes]?

Formant transitions

  • F1 always rises after a stop.

http://www.cns.nyu.edu/ david/courses/perception/lecturenotes/speech/speech.html

Formant transitions

  • The direction of F2 and F3 formant transitions depend
  • n the particular constriction producing the stop (lips,

tongue tip, tongue body).

http://www.cns.nyu.edu/ david/courses/perception/lecturenotes/speech/speech.html

slide-8
SLIDE 8

Formant transitions

  • Formant transitions at the start of the vowel following a

voiced stop (reverse for vowels preceding voiced stop):

F2 F3 [b] rises rises [d] ∼level falls [g] falls rises Formant transitions

  • Formant transitions at the start of the vowel following a

voiced stop (reverse for vowels preceding voiced stop):

From Ladefoged (2001)

ebb vs. egg

ebb egg

1000 2000 3000 4000 5000 1000 2000 3000 4000 5000

vowel stop closure

release

vowel stop closure

release

Velar stops [k] [g]

  • Velars are relatively easy to spot due to the ‘velar pinch’
  • f F2 and F3.
slide-9
SLIDE 9

Velar stops [k] [g]

  • Formant transitions for [g] tend to be longer for other

stops – tongue body is also involved in vowel formation.

Velar stops [k] [g]

  • The place of constriction for velar stops can vary – why?

Voiceless stops [p] [t] [k]

  • In voiceless stops, the movement of lips and tongue take

place during the burst, prior to the onset of voicing.

  • Therefore these movements are not as clearly reflected

in formant transitions/spectrograms

  • A more reliable cue to the place feature for voiceless

stops is the noise burst.

Voiceless stops [p] [t] [k]

  • In voiceless stops, the movement of lips and tongue take

place during the burst, prior to the onset of voicing.

  • Therefore these movements are not as clearly reflected

in formant transitions/spectrograms

  • A more reliable cue to the place feature for voiceless

stops is the noise burst.

slide-10
SLIDE 10

Voiceless stops [p] [t] [k]

  • In voiceless stops, the movement of lips and tongue take

place during the burst, prior to the onset of voicing.

  • Therefore these movements are not as clearly reflected

in formant transitions/spectrograms

  • A more reliable cue to the place feature for voiceless

stops is the noise burst.

Voiceless stops [p] [t] [k]

  • The closure is devoid of voicing bar activity...

Time (s) 0.738 5000

Voiceless stops [p] [t] [k]

  • ...and the burst can be clearly seen in the spectrogram.

Time (s) 0.738 5000

Voiceless stops [p] [t] [k]

  • However, not all bursts are created equal.

Time (s) 0.738 5000

slide-11
SLIDE 11

Energy location Energy strength Length of burst [p] wide range faint short [t] mostly > 4 kHz strong long [k] mostly < 4 kHz strong longest k t p

Frequency

From Kent & Read (1992), The Acoustic Analysis of Speech

Voiced vs. voiceless stops

aba

Voiced vs. voiceless stops

apha

Voiced vs. voiceless stops

aba apha

slide-12
SLIDE 12

Voiced vs. voiceless stops

[kAA] ‘neck’

Time (s) 0.4504 5000 Frequency (Hz) Frequency (Hz)

[khAA] ‘donkey’

Time (s) 0.4566 5000 Frequency (Hz) Frequency (Hz)

Other cues

Frequency (Hz) Time (ms) 5000 5000

/gʌk/ /gʌg/

  • closure duration
  • voicing during closure
  • duration of preceding vowel

Other cues

Frequency (Hz) Time (ms) 5000 5000

/gʌk/ /gʌg/

  • closure duration
  • voicing during closure
  • duration of preceding vowel

Other cues

Frequency (Hz) Time (ms) 5000 5000

/gʌk/ /gʌg/

  • closure duration
  • voicing during closure
  • duration of preceding vowel
slide-13
SLIDE 13

Other cues

Frequency (Hz) Time (ms) 5000 5000

/gʌk/ /gʌg/

  • closure duration
  • voicing during closure
  • duration of preceding vowel

Spectrogram practice

  • Ladefoged course website:

http: //www.phonetics.ucla.edu/course/chapter8/figure8.html

  • More spectrograms:

http://home.cc.umanitoba.ca/~robh/howto.html

Spectrogram practice

  • Ladefoged course website:

http: //www.phonetics.ucla.edu/course/chapter8/figure8.html

  • More spectrograms:

http://home.cc.umanitoba.ca/~robh/howto.html