Part 1: Acoustic Phonetics Data Flashcards
Used by many disciplines
Linguists
Speech communication specialists–speech synthesis & recognition systems
Speech production scientists
Speech perception scientists
Speech-language pathologists
V=vowel
/h/ has minimal
əhVd” frame is commonly used
V=vowel
/h/ has minimal influence on the vocal tract gestures required for a following vowel
/d/ provides a “natural” ending to the syllable
Accommodates the production of lax vowels such as /ɪ/, /ε/, and /ʊ/
Vowel Formants
a single vowel can be associated
a single vowel can be associated with a wide range of F1-F2 values depending on resonance patterns of tubes of different lengths, and age- and sex-related differences in vocal tract length
In cases of overlapping formant freqs, identity of the speaker, including age, sex, and dialect, allows listeners to link a specific formant pattern to a vowel category
Vowel Quadrilaterals for men, women, and children
quadrilVowel aterals for men, women, and children move from the lower left to upper right part of the graph, respectively.
Vocal tract becomes progressively shorter
Vowel space appears to be larger for children, compared with men and larger for women compared with men
the vowel quadrilateral for one group of speakers cannot be perfectly fit to the quadrilateral for a different group by moving it to the new location
Vowel Spaces
Acoustic vowel space for corner vowels has
Acoustic vowel space for corner vowels has clinical application as an index of speech motor integrity
E.g., smaller in persons with dysarthria
Size of the acoustic vowel space is correlated with speech intelligibility or perceptual measures of articulatory precision
Can be made to expand and contract with different speaking styles
Interpretations of the correlation between Vowel Spaces & Speech Intelligibility
Size (area) of the vowel space may be
Size (area) of the vowel space may be an index of articulatory mobility and speech motor control
Larger vowel space areas increase the acoustic difference between closely related vowels, such as /i/ versus /ɪ/ or /u/ versus /ʊ/
independent component of a speech intelligibility deficit (separate from motor control)
Branches of Comparative Acoustic Phonetics
Acoustic characteristics of similar speech sounds in two or more
Acoustic characteristics of similar speech sounds in two or more languages or in two or more dialects of the same language.
The effect of native language (or dialect) phonetics on the acoustic characteristics of speech sounds in a second language
Acoustic characteristics of vowels have been a major focus for both branches
Within-speaker variability in formant frequencies
vowel formant frequencies vary with a number of factors
vowel formant frequencies vary with a number of factors
speaking rate
syllable stress
speaking style
phonetic context
Articulatory Undershoot
Lindblom (1963) coined the term to describe
Lindblom (1963) coined the term to describe vowel production in connected speech that was not the most extreme configuration associated with the sound
the shorter the vowel duration, the greater the undershoot
increased speaking rate, reduced stress, and a casual speaking style are all associated with shorter vowel durations
Summary of Vowel Formant Frequencies
Sex and age impact
Sex and age impact the formant frequencies of vowels because they are related to differences in vocal tract size and length.
In general, the longer and larger the human vocal tract, the lower the formant frequencies for all vowels
explains the large range of formant frequencies across the population
Even when vocal tract length/size factors are held
Even when vocal tract length/size factors are held constant, “target” formant frequencies for a given vowel may vary for several reasons
Dialect, vocal tract length, phonetic context, syllable stress, speaking rate, speaking style
Vowel Durations
Studied extensively due to application to:
Studied extensively due to application to:
speech synthesis
machine recognition of speech
description and possibly diagnosis of certain speech disorders
Extrinsic Factors Affecting Vowel Durations
Consonant voicing
Vowels are typically longer when surrounded by voiced consonants
Stress
Vowels in emphasized syllables have greater duration
Speaking rate
Slower rates=longer vowel durations and vice versa
Extrinsic Factors Affecting Vowel Durations
Utterance position
Utterance position
Phrase-final or utterance-final lengthening
Speaking style
E.g., “Clear speech”
Diphthongs
The six diphthongs in American English include:
/ɑɪ/ (“guys”)
/ↄɪ/ (“boys”)
/ɑʊ/ (“doubt”)
/eɪ/ (“bays”)
/oʊ/ (“goes”)
*/ju/ (“beauty”)
*not considered a diphthong in many phonetics textbooks, but has properties similar to the other diphthongs.
Diphthongs: Two connected vowels or a unique phoneme?
absence of steady states in diphthongs is a potential
absence of steady states in diphthongs is a potential complication in classifying diphthongs as a sequence of two vowels
each of the diphthongs has an identifiable transitional segment (as seen on spectrograms
Diphthong Duration
Diphthongs are generally longer than monophthong vowels in equivalent environments and speaking conditions
Nasal articulations are described
Nasal articulations are described acoustically in two categories
Nasal murmur
Nasalization
Three factors allow human listeners or statistical classification to
Three factors allow human listeners or statistical classification to allow fairly accurate identification
murmur offset
vowel onset (murmur + transition piece)
transition piece
Nasalization
Involves complex acoustics resulting from
Involves complex acoustics resulting from the mix of oral and nasal tract formants with antiresonances originating in the sinus cavities
Fourier spectrum of vowels
A1—P1 relative amplitude difference
Nasalance
Measure of the acoustic energy radiating from the nares
Nasometry values are often computed for extended passages.
One with no nasal consonants
One loaded with nasals to elicit high nasalance values
May use a passage with a mix of obstruents and nasals is estimate nasalance for the phonetics of typical utterances.
Advantages of Nasometry
speed of obtaining a value
automatic nature of the measurement
tendency for perceptual estimates of nasality to increase as nasalance increases
large number of published articles on nasalance values in various populations
Disadvantages of Nasometry
global nature of the measure
usually averaged across entire passages and yields a single number per passage
tendency for perceptual estimates of nasality to vary imperfectly with nasalance values
difficulty of knowing how variation in nasalance values relates to the specific nature of velopharyngeal dysfunction
Semivowels
/w/, /ɹ/, /l/, and /j/
require movement to and away from a vocal tract constriction tighter than that for vowels, but not as much as for obstruents
All are produced with a vocal tract open to the atmosphere, and are considered vocalics.
/w/ and /j/ are also referred to as glides, /ɹ/ and /l/ as liquids
constriction interval and transition acoustics provide the acoustic information necessary to distinguish among these sounds.
Constriction Interval
interval of relatively “flat” formants
assumed to correspond to the part of semivowel articulation when the vocal tract is most constricted
formant pattern like those of vowels
Formant Transitions
pattern of formant transitions into and out of the constriction intervals also distinguishes among the semivowels
Important characteristics (see 11-9)
the specific formants that have large transitions into and out of the constriction interval
the direction (rising versus falling) of the transitions
Semivowel Acoustics and Speech Development
semivowel errors are
semivowel errors are frequent during phonological development and in speech delay
E.g., /w/ for /ɹ/, /w/ for /l/, and /j/ for /l/
Need to determine if the issue is due to articulatory control needed to differentiate the sounds or distinguishing the perceptual representations
Semivowel Acoustics and Speech Development
the acoustics of a [w] in a [w] for /ɹ/ error
the acoustics of a [w] in a [w] for /ɹ/ error (or any other substitution error) are often not like the acoustics of normally articulated [w]
the error [w] is different from correct [w] by having acoustic characteristics more or less between the error sound and the correct sound
This shows that the child hears the difference but has difficulty with articulation
A distinction is made by the child but may be too subtle for human listeners to perceive.
Even if listeners do hear a subtle distinction they may place it in a “comfortable” phoneme category
Semivowel Durations
Challenging to segment
Challenging to segment semivowels from adjacent vowels
When constriction intervals can be segmented from the surrounding transitions they have durations of 30 to 70 ms, with the majority of values toward the lower end of this range
Semivowel Durations
Combined duration of
Combined duration of the transition and constriction intervals of semivowels may be brief (as short as 100 ms)
Suggests rapid, complex articulatory gestures occurring in a short amount of time–may explain, in part, why children master the contrasts of these sounds relatively late in the overall scheme of phonological development
Fricatives
Characterized by an interval of aperiodic energy whose spectrum and overall amplitude depend on place of articulation and, in some cases, voicing status.
In English, fricatives are categorized as sibilants (/s, z,ʃ,ʒ/), nonsibilants (/f, v,θ,ð/), and the glottal fricative /h/
Fricatives
Sibilants are more
Sibilants are more intense and have better-defined spectra than nonsibilants.
Sibilants have more easily identified spectral peaks and concentrations of spectral energy
Sibilants vs. Nonsibilants: Spectral Characteristics
The intensity difference is represented in the spectrogram by the much darker frication noise
Overall higher level of /s/ compared with /f/
This intensity difference is consistent for any sibilant-nonsibilant comparison
Higher intensity for sibilants is largely due to an obstacle (i.e., teeth) in the path of the airstream
Sibilants typically have peakier spectra than nonsibilants
/f/ spectrum is flatter than the /s/ spectrum
Quantification of Fricative Spectra
The “peak frequency” does not consider additional information in fricative spectra.
Varying shapes
Spectral moments: four numbers that represent the spectral shape
basic statistical properties of a distribution of numbers, applied to speech-sound spectra
Formant Transitions and Fricative Distinctions
Evidence for the importance of
Evidence for the importance of formant transitions in distinguishing place of articulation for nonsibilant fricatives is mixed.
Jongman et al. (2000) failed to identify formant transition patterns that consistently separated the four places of fricative articulation
Fricative Duration
voiceless fricatives are longer
voiceless fricatives are longer than voiced fricatives
Other factors can influence duration
Position-in-word
Stress level of the syllable
Speaking rate
Phonetic context
Sibilants are longer than nonsibilants
Laryngeal Devoicing Gesture and Fricative Duration
Voiceless fricatives require the laryngeal devoicing gesture (LDG), an opening-closing movement of the vocal folds observed for voiceless obstruents
The opening plus closing motions of the LDG produce a very long event compared to the short-duration opening and closing motions of a single cycle of vocal fold vibration.
The LDG typically has a duration of roughly 120 to 150 ms, whereas an example of a “long” period for one phonatory cycle would be roughly 10 ms
/h/ Acoustics
glottis is partially or largely open and the aperiodic source is produced when the air jet emerging from the constriction between the vocal folds strikes the edges of the ventricular folds and epiglottis, generating turbulent airflow
Source-Filter Acoustics for /h/
/h/-interval contains aperiodic energy
Intervocalic /h/ often has a combination of aperiodic and weak, periodic energy
suggesting that the abducted vocal folds are vibrating loosely, with minimal or absent closed phases
Source-Filter Acoustics for /h/
Formants can be detected during the /h/-intervals
Formants can be detected during the /h/-intervals
/h/ is produced with the vocal tract shape of the surrounding vowel(s)
Relative weakness of energy around F1 compared with the much more intense energy of the upper formants
Due to sound absorption in the trachea
Loss of F1 energy resulting from vocal fold vibration in which there is poor closure
Explains intelligibility problems due to indistinct vowels in people with breathy voice
Stop Consonants
Only consonant type to occur in virtually all languages of the world
Among the most frequently occurring consonant segments
Closure Interval and Burst
Closure interval and burst are acoustic markers of stop consonants
Closure interval corresponds to the interval during which the vocal tract is completely sealed
Closure intervals for voiceless stops appear as white intervals on spectrograms
Closure interval is white with periodic energy along the baseline indicating vibration of the vocal folds for voiced stops
Closure Interval Duration
In connected speech, closure intervals for any
In connected speech, closure intervals for any of the stop consonants are rarely greater than 70 ms, the majority having durations of around 60 ms
In more formal utterances, the values are often longer than 70 ms but rarely exceed 100 ms
Closure Interval Duration
Claims that stop closure durations
Claims that stop closure durations are longer for voiceless compared with voiced stops
Claims that durations become increasingly shorter as place of articulation moves back in the vocal tract
(i.e., /p/ closures are longer than /t/, and, are longer than /k/)
Inconsistency in research regarding stop closure durations (Table 11-6)
Closure Interval Duration
Voiceless stop closure durations between 2 and 6
Voiceless stop closure durations between 2 and 6 ms longer than voiced stops in connected speech
Difference of insufficient magnitude to serve as a useful perceptual cue to the voicing status of a stop.
Flap Closures
t/ and /d/ in the intervocalic, post-stressed position of words (e.g., butter, matter, ruder)
Shorter than stop closures in more formal speech styles
Closure Duration and Place of Articulation
Tendency for labials to have the longest durations
Differences between lingua-alveolar and dorsal closure durations are less clear
Stop Voicing: Some Further Considerations
Acoustics of stop consonant voicing are
Acoustics of stop consonant voicing are complex
Cues that signal the voicing status of a stop are context dependent
absence of glottal pulses during a voiceless closure interval is almost always the case, but glottal pulses do not always occur throughout the entire closure interval of a voiced stop
Stop Voicing: Some Further Considerations
Without other potential cues for the
Without other potential cues for the stop voicing distinction, voiced stops may be considered more vulnerable to perceptual errors compared with voiceless stops
Speakers with speech motor control problems (e.g., apraxia, dysarthria) more often produce voiceless-for-voiced than voiced-for-voiceless errors
Laryngeal Devoicing Gesture, Stop Closures, and Voice Onset Time
LDG explains
LDG explains the absence of glottal pulses within the stop closure interval
Synchronized onsets of the LDG and supralaryngeal closure
LDG duration of well over 100 ms for stops (see Figure 11-29)
Release of the stop consonant closure interval roughly 60 to 70 ms after its onset (compared with the fricative constriction over the entire duration of the LDG )
VOT and Voiced Stops
Voiced stops do not have an LDG
During the closure interval, the vocal folds remain in the midline, phonation-ready position
The value of VOT is grossly correlated with the presence versus absence of glottal pulses within a stop closure interval.
positive and negative VOT values are common for voiced stops
VOT and Voicing Status of Stop Consonants
Value of VOT can generally be used as a correlate of the voicing status of a stop consonant
“Mismatches” between VOT values and the voicing status for stops usually occur only when voiceless stops have a VOT in the short-lag range.
Voiceless stops in the post-stressed position of a word
Voiceless stop is part of an s + stop cluster within a single syllable
VOT Across Languages
Many languages have stop voicing distinctions, but divide the VOT range in different ways compared with English.
different languages exploit the VOT continuum to “implement” their unique voicing contrasts
Korean has a three-way voicing distinction for each place of articulation
French has a two-way voicing contrast but divides the VOT continuum differently than English
Voiced stops of French all have negative VOTs (i.e., they are prevoiced), and short-lag, positive VOTs for voiceless stops
Bursts
Spectra for burst sources
Spectra for burst sources and frication sources are similar even though their aeromechanics are different
Both source spectra are shaped significantly by the resonator in front of the source
Bursts
Considered one of the hallmarks
Considered one of the hallmarks of the stop manner of production
However, many stop consonants have no identifiable burst in connected speech
Halle et al. (1957) Study of Spectral Bursts
Labial bursts: primary
Labial bursts: primary concentration of energy in the lower frequencies
Lingua-alveolar: flat spectra or an emphasis of energy in the high frequencies (above 4.0 kHz)
Dorsal stops: prominent energy peaks in the midfrequency regions, roughly between 1.5 and 4.0 kHz.
Acoustically Invariant Characteristics
Characteristics that are always found in
Characteristics that are always found in the spectrum, or the manner in which the spectrum changes as a function of time, regardless of who is producing the utterance or under what conditions the utterance is spoken.
complicated by the phenomenon of coarticulation
Coarticulation
Influence of one segment on another
Articulatory (and acoustic) characteristics of a speech sound segment depend on the articulatory (acoustic) characteristics of adjacent, and in some cases nonadjacent, segments
Acoustic Invariance for Stop Place of Articulation
Vowel-induced variation could result in highly variable stop burst spectra for a given place of articulation
Too little acoustic stability to serve as a reliable cue to place identification.
Burst Spectra and Acoustic Invariance
The auditory system may strip
The auditory system may strip away spectral detail and use the gross spectral shape to identify stop place of articulation.
Blumstein and Stevens (1979) described spectral shapes for the three different places of stop articulation.
diffuse-falling for bilabials
diffuse-rising for lingua-alveolars
compact for dorsals
Acoustic Invariance and Theories of Speech Perception
Early perceptual experiments
Early perceptual experiments resulted in a conclusion of “no acoustic invariance” for stop consonants
Haskins scientists found they could elicit perception of a stop-vowel syllable with a pattern having only F1-F2 transitions and the steady states of the following vowel
burst was not necessary for perception of a stop consonant
Acoustic Characteristics Related to Place of Articulation of a Stop
Because vowel context had
Because vowel context had such influence on acoustic characteristics of stop consonants, Haskins scientists argued that there was no acoustic invariance for the place feature of stop consonants.
Led to a theory of speech perception that downplays (if not eliminates) a primary role for auditory acoustic analysis in the perception of speech
Arguments for Acoustic Invariance
Blumstein and Stevens (1979
Blumstein and Stevens (1979), and others, had success in classifying stop place of articulation using the burst spectrum
Ability to assign 85% to 95% of burst spectra to the “correct” place of articulation pointed to sufficient acoustic invariance for stop consonants
Locus Equations
Regression equations that combines measures reflecting both consonant and vowel production
Developed by Sussman, Lindblom et. al
Acoustic Invariance at the Interface of Speech Production and Perception
Acoustic characteristics vary for many reasons related to the speaker and conditions
Acoustic characteristics vary for many reasons related to the speaker and conditions
Age
Dialect
Sex
Size
Phonetic context
Speaking rate
Speaking style
If the acoustic characteristics of a speech sound are so variable, exactly what is meant by the term “acoustic invariance”?
Interpretations of Acoustic Invariance
Haskins scientists had a
Haskins scientists had a stricter interpretations
virtually no acoustic variability
Blumstein & Stevens (1979) and others had a looser approach
Allowed that a sufficient degree of acoustic invariance was demonstrated according to their analysis criteria
Lindblom (1990) said the issue is how much acoustic characteristics can vary and still remain distinctive relative to neighboring sounds.
Affricates
Frication interval of the English affricates /tʃ/ and /d3/ is longer than the frication interval of the stop component and shorter than the typical duration of the fricative component
Stop closure duration of affricates tends to be slightly shorter than the closure duration of singleton stops
Affricates
Place of articulation of the English
Place of articulation of the English affricates is slightly posterior to the place of articulation for lingua-alveolar stops and fricatives
Place of articulation of the English affricates is slightly posterior to the place of articulation for lingua-alveolar stops and fricatives
Acoustic Characteristics of Prosody
Intonation
Rhythm
Stress
Pause
Grammatical function
Phrase-Level F0 Contours
For declarative phrases, an F0 peak typically occurs near the beginning of the phrase, followed by a gradually declining F0 to the lowest value at the end of the phrase
An interrogative F0 contour typically has a rising F0 at the end of an utterance
Phrase-Level F0 Contours
Utterance-final rise of F0 is
Utterance-final rise of F0 is a cue to the listener, along with other cues (lexical, syntactic, situational), that a question is being asked
F0 contours may also rise at the end of a declarative utterance when the speaker intends to continue talking (less than rise for questions)
F0 contours may also be modified by the level of stress produced on a specific syllable
Paralinguistics
Nonverbal aspects of speech, usually related to prosodic variation, that convey emotion and physiological status.
Can be thought of as a backdrop against which a message is transmitted.
Phrase-Level Intensity Contours
In connected speech, an intensity contour varies over time primarily because consonants are less intense than vowels
7 to 14 dB intensity differences between vowels and consonants
Intensity differences between consonants and vowels depend on many factors
Type of vowel
Consonant word position
Syllable stress level
Overall vocal effort level
Sex of the speaker
Stress
English has multisyllabic words in which one syllable (sometimes two) is (are) stressed relative to the others.
Lexically stressed syllables
higher F0
greater intensity
longer duration
possibly a less reduced vowel formant pattern compared with unstressed syllables
Stress
Prominence–refers to
Prominence–refers to syllables that “stand out” for a listener
Often marked by intensity and duration, with F0 playing a weak role
Clinical application
Patients who have difficulty stressing syllables for lexical or sentence stress have multiple ways to make a syllable prominent
Rhythm
Patterning of segment or syllable durations across an utterance
English is said to have a rhythm in which long and short-duration syllables alternate with each other.
The long syllables are the stressed ones, the short syllables are unstressed.
Usually a sequence of one long followed by several short syllables
Rhythm
English is considered a
English is considered a stress-timed language
requires the duration between stressed syllables in an utterance to be nearly constant.
In Spanish, duration between consecutive syllables, rather than stressed syllables, is nearly constant
Rhythmic Abnormalities
Certain speech disorders exhibit rhythmic abnormalities
English speakers with cerebellar disease may produce speech with a rhythm that distorts the normal approximation to stress timing by making all syllables roughly equal in duration.
Pairwise Variability Index (PVI)
acoustic measure of the relative duration difference between consecutive syllables in connected speech.
Has potential as a diagnostic marker for different types of speech disorder
Review
Acoustic phonetic analysis
Acoustic phonetic analysis is useful for SLPs to gain insight into an individual’s speech production problems
Concepts are relevant for AuD to understand the relevance to speech intelligibility and hearing devices
Review
Acoustic characteristics
Acoustic characteristics of vowels, diphthongs, nasals, semivowels, fricatives, stops, and affricates are described with respect to formant frequencies, antiresonances, formant transitions, spectral shapes, segment durations, and segment intensities
Review
F0 and intensity contours
F0 and intensity contours are significant prosodic characteristics
Acoustic characteristics of lexical and sentence stress are variable across speakers
