READING SPECTROGRAMS PhonPhon 1 Acoustic Phonetics Overview Structure

READING SPECTROGRAMS Phon-Phon 1

Acoustic Phonetics: Overview • Structure of sound: simple periodic sounds • Structure of speech sound: • Complex periodic sounds • Aperiodic sounds • Generating speech sounds • Source-Filter theory • Formants • Measuring speech sounds • Wave forms • Spectrograms 2

Recall: Key Properties of Sound Waves property waveform spectrum amplitude time (period) y-axis x-axis y-axis NONE frequency calculate x-axis

Visualizing Properties of Sound Waves property amplitude time (period) waveform spectrum spectrogram y-axis shading x-axis NONE x-axis frequency calculate x-axis y-axis

Spectrogram with Waveform Note: Praat software gives spectrogram with waveform as default Waveform Spectrogram

(lig. d vs e k) ar pli tu ht d Am Frequency Spectrograms Time

Formants for different vowels

Formants for different vowels

Formants for different vowels

Observing the acoustics of consonants • We will start with sonorant consonants. • Why? • Same sound source as vowels.

Observing the acoustics of consonants

Sonorant Consonants • Which consonants? • Approximants • glides [w], [j] • liquids [l], [ɹ] • Nasals • [m], [n], [ŋ] • Articulatory Phonetics • Consonants = voice, place, and manner

Sonorant Consonants • Voice • All voiced • Complex periodic sounds adjusted by vocal tract like vowels • more constricted than vowels • we will see formants! • Place • Reflected in what happens to the formants • Manner • Reflected in what happens to the formants

Sonorant Consonants: Glides [aja] • F 1 low • F 2 high [aia] ◦ F 1 low ◦ F 2 high

Sonorant Consonants: Glides [awa] • F 1 low • F 2 low [aua] ◦ F 1 low ◦ F 2 low

Acoustics of Glides • [j] ~ [i] and [w] ~ [u] • Similar formant measurements • Glides shorter and quieter • Why does this make sense? • /j/ = palatal approximant • constriction near palate like front vowels • greater constriction than vowels = very high • /w/ = labio-velar approximant • constriction closer to velum like back vowels • lip rounding like /u/ • greater constriction than vowels = very high

Sonorant Consonants: Liquids • Laterals (like /l/) and rhotics (like /ɹ/) • Among the most sonorous consonants • Can be syllabic • e. g. , ladle [leɪdl ] dapper [dæpɹ ] / [dæpɚ] • Why are they grouped together? • Often pattern together phonotactically • e. g. , may form a cluster • [pɹV] or [pl. V] • [Vɹp], [Vlp] • Other languages: • allophonic relationship • e. g. , Korean: rhotic in onset, lateral in coda • in free variation (can use either)

Sonorant Consonants: Liquids • Acoustics: • Resemble vowels with lowered intensity (like glides) • Both /l/ and /ɹ/ have low F 2 (similar to /w/ glide) • F 3 lowered dramatically in /ɹ/ but not /l/ / ɹaɪ/ Source: Idemaru & Holt (2009) /laɪ/

Sonorant Consonants: Approximants

Sonorant Consonants: Nasals • Sealed-off Oral Cavity: The Implications • Anti-resonances • frequencies at which amplitudes of the harmonics are severely reduced • happens when air in the vocal tract becomes divided between oral and nasal cavities • some sound energy trapped in oral cavity and not resonated through nasal passages • Hence we see anti-formants • area of unusually low intensity in the spectrogram (blank space)

Sonorant Consonants: Nasals • Access to nasal cavity: The Implications • Nasal “murmur” or nasal formant • with nasal cavity, the vocal tract becomes longer • longer resonator = lower frequencies • very low formant (nasal murmur) visible on spectrogram (200 -300 Hz for males) • lower for [m] than [n]; lower for [n] than [ŋ] • [m] makes the biggest VT • [ŋ] the smallest

Acoustics of Nasals decreased amplitude abrupt change antiformant F 2 rising out of bilabial low F 1 [ɑ] [m] [ɑ]

Sonorant Consonants: Nasals • Voiced = formants appear • lower intensity • Place = measurement of first anti-formant • lowest for /m/ • highest for /ŋ/ • Manner • low nasal murmur (nasal formant) • anti-formants • low intensity compared to other sonorants

Obstruent Consonants

Voiceless Obstruents • Stops/plosives • /p/, /t/, /k/ • Fricatives • /f/, /θ/, /s/, /ʃ/, /h/ • Affricates • /ʧ/ • All have voiced counterparts (except /h/)

Voiceless Obstruents: Voice • none of the voiceless obstruents are voiced! • sound source is not phonation • no complex periodic sound wave from the vocal folds • no formants! • sound source = air escaping constriction in the upper vocal tract • creates turbulence • generates noise (= aperiodic sound) • Resonance? • still happens • most effective resonators are those in front of place of constriction • resonator is still the vocal tract, but the parts that resonate are those in front of construction • most relevant for fricatives

(Voiceless) Obstruents: Manner & Place • Manner • Fricatives = continuous turbulent airflow • Stops = complete cessation of airflow + turbulent release • Affricates = stop closure is released into fricative airflow • Different acoustic correlates for each of these • Place • Cues for place will vary with manner

Fricatives: Articulation • Articulators brought close together but without complete occlusion • Require precision • Must have narrow enough constriction to generate frication (turbulent air) • But not so great a constriction that air is fully stopped • English: / f, v, θ, ð, s, z, ʃ, ʒ, h /

Acoustics of Fricatives • Sound source • Frication noise generated by turbulence at point of constriction • at teeth for /θ/, alveolar ridge for /s/, etc. • Resonator • Portion of vocal tract in front of constriction

Spectrogram: “see” can’t really see formants for [s] clear formants for [i]

Waveform: “see” Aperiodic Sound Source for [s]

Waveform: “see” Periodic Sound Source for [i]

Fricatives: Sound Source • Two types of turbulence • Channel turbulence • noise produced when airflow escapes from a narrow channel and hits inert outside air • Obstacle turbulence • noise produced when airflow hits an obstacle • louder than channel turbulence

Channel vs. Obstacle Turbulence /f/ and /θ/ • not a lot of obstacles to bounce off of • rely more on channel turbulence • quieter /f /θ / /

Channel vs. Obstacle Turbulence /s/ and /ʃ/ • tongue grooves in order to aim airflow directly at teeth • teeth = obstacle turbulence • louder /s /ʃ / /

Fricatives: Resonance • Resonance of fricatives depends on the length of the tube (“filter”) in front of the constriction (“source”) • Backer constriction lower resonant frequencies • Fronter constriction higher resonant frequencies

Fricatives: Length of cavity in front of constriction

Fricatives: • Resonance of fricatives depends on the length of the tube (“filter”) in front of the constriction (“source”) • Backer constriction lower resonant frequencies • Fronter constriction higher resonant frequencies • helps us differentiate /s/ vs. /ʃ/ • doesn’t work so well for /f/ vs. /θ/ • identical place of articulation (upper teeth) • different active articulator (lips vs. tongue) • not a lot of noise to begin with (flat spectrum) • can try to look at following vowel for cues to place (formant transitions more labial in /f/, more coronal /θ/) • but we haven’t learned this yet • and it’s still hard

Sibilants • [s] and [ʃ] are known as sibilant fricatives • more acoustic energy at higher frequencies than other fricatives because they: • are obstacle fricatives • air aimed at teeth louder • have small, short resonating cavity higher frequency resonance Sibilants Not sibilants

Fricatives: Place • Wait a minute… what about [h]? • look, three clear, neat peaks… like a vowel!

/h/ • Basically a voiceless vowel • takes on (very weak) formants of whatever vowel follows it • source at glottis, like vowel, but aperiodic • vocal tract already in place for next vowel

Voiceless Fricatives: Manner • Turbulent release of air = broad range of frequencies and no real formants [s] [f] [ʃ] [θ] [h] "Some find shiny things here. ” http: //home. cc. umanitoba. ca/~robh/archives/arc 0703. html

Fricatives: Place • Place revealed in intensity at different frequencies • [f], [θ]: less intense, even distribution of intensity, harder to tell apart • [s], [ʃ]: high intensity, [ʃ] has lower intense frequencies • [h]: can sometimes see weak formants for following vowel [s] [f] [ʃ] [θ] [h] "Some find shiny things here. ” http: //home. cc. umanitoba. ca/~robh/archives/arc 0703. html

Voiceless Fricatives: Summary • Turbulence provides the source of fricative noise • obstacle turbulence ([s], [ʃ]) louder than channel turbulence ([f], [θ]) • Filter of fricative turbulence noise is based on the place of articulation • [f], [θ]: essentially no filter (flat spectrum) • [s], [ʃ]: short filter, emphasizing higher frequencies • [h]: longer, more vowel-like filter • Sibilants are loud fricatives with concentrations of intensity at high frequencies • [s], [ʃ]

(Voiceless) Obstruents: Manner & Place • Manner • Fricatives = continuous turbulent airflow • Stops = complete cessation of airflow + turbulent release • Affricates = stop closure is released into fricative airflow • Different acoustic correlates for each of these • Place • Cues for place will vary with manner

Stops: Articulation • Three stages of articulation • Closing phase • Maintain closure • Release of closure

Stops: Articulation & Acoustics • Maintain closure • acoustically, nothing is happening • completely blank space on spectrogram • Release of closure • “release burst” of air • vertical “spike” of energy on waveform • more prominent for voiceless than voiced stops

Voiceless Stops: Manner • First, silence • blank space on spectrogram • (maybe some residual voicing, but not much) • Then, release burst • Air rushes out of the mouth at high speed, producing a pressure impulse that lasts ~ 2 -3 ms • “transient” • produced at place of articulation • i. e. , the source during the release is the air that escapes from that place of articulation

[ata] release burst silent gap

[ada] release burst silent gap

Stops: Place • Formant Transitions • change in formants in the vowels on either side of a consonant • reflects place of articulation • F 1 • always low next to stops • (drops into stops, rises out of stops) • high tongue position = low F 1 • stop = highest tongue position there is (full closure) • F 2 • place of constriction shapes transitions for stops • direction of formant movement depends on place of consonant and the vowel

Oral Cavity Size and F 2 small oral cavity high F 2 large oral cavity low F 2

Formant Transitions: Labials F 2 F 1 Labials • largest oral cavity possible • larger than any vowel • lower F 2 than any vowel • F 2 goes from low to high into a vowel • F 2 goes from high to low into a labial

Formant Transitions: Coronals • larger oral cavity than frontish vowels • F 2 lower than vowel • smaller oral cavity than backish vowels • F 2 higher than vowel

Formant Transitions: Dorsals • smallest oral cavity possible • smaller than any vowel • higher F 2 than any vowel • F 2 goes from high to low into a vowel • F 2 goes from low to high into a velar

Formant Transitions: Velar pinch Second and third formants approach each other in velars [bag] [bak]

Obstruents: Voiced vs. Voiceless • Voiced obstruents • Sound source = phonation and VT-generated noise • Fricatives • voicing bar (low frequency) • voiced versions also usually shorter and quieter • Stops/plosives • voicing bar (low frequency) • VOT (Voice Onset Time) in English

Voiced fricatives: Waveform & Spectrogram [-voice] [+voice] aperiodic no voice bar

Voiced Fricatives • How can we tell the difference between voiced and voiceless fricatives on a spectrogram? • voicing bar! • How about a waveform? • voiced fricative shows aperiodic noise superimposed on periodic signal of vocal fold vibration Source: http: //home. cc. umanitoba. ca/~krussll/phonetics/acoustic/sp Source: http: //clas. mq. edu. au/acoustics/waveforms/speech_waveforms. ht

Stops: English Voicing Contrast • Phonologically: • “voiceless” vs. “voiced” • Phonetically: • In initial position (usually) • /p/: Voiceless aspirated [pʰ] • /b/: Voiceless unaspirated [p] • /sp/: Voiceless unaspirated [p] (same as initial /b/) • Intervocalic • /p/: Voiceless aspirated [pʰ] • /b/: Voiced [b] (sometimes)

Stops: Voice Onset Time (VOT) • Time between the release of a stop closure and the beginning of voicing (usually for a vowel) • On waveform/spectrogram • milliseconds as the duration between the vertical spike marking transient burst and the first vocal pulse that can be observed at baseline • Positive: voicing starts after stop release • Negative: voicing starts before stop release

Stops: Voice Onset Time (VOT) • Time between the release of a stop closure and the beginning of voicing (usually for a vowel) • For aspirated stops: 1. Stop closure is made 2. Airflow builds up pressure behind closure 3. Closure is released (with a “burst”) 4. Air flows unimpeded through glottis (“aspiration”) • different source than release burst! • burst = area of constriction • aspiration = glottis 5. Vocal folds adduct & voicing begins for vowel

Stops: Voice Onset Time (VOT) • time between the release of a stop closure and the beginning of voicing (usually for a vowel) • For unaspirated stops: 1. Stop closure is made 2. Airflow builds up pressure behind closure 3. Closure is released (with a “burst”) 4. Air flows unimpeded through glottis (“aspiration”) 5. Vocal folds adduct & voicing begins for vowel AT THE SAME TIME AS (3)

Stops: Voice Onset Time (VOT) Note: A true “voiced” plosive would adduct the vocal folds earlier (during hold)

Stops: Acoustic cues for “voicing” • Voiceless aspirated stops: • VOT 40 - 100 milliseconds • Voiceless unaspirated stops: • VOT 0 - 20 milliseconds • (these are our “voiced” stops in English) • Voiced stops: • VOT < 0 milliseconds • voicing already occurs during closure • uncommon in American English

“peach” 57 ms VOT

“beach” no voice bar 7 ms VOT

“speech” 7 ms VOT

[ɑʤɑ] frication silent gap

[ɑʤɑ] periodicity! voice bar

A Grand Summary: Manner

A Grand Summary: Place

A Grand Summary: Voice

Practice: Segmenting Waveforms

Practice: Segmenting Spectrograms