- Количество слайдов: 31
CSE 551: Structure of Spoken Language Lecture 5: Characteristics of Manner of Articulation John-Paul Hosom Fall 2004
Self-Study If you want to look at spectrograms of your own voice, there are several programs available: 1. Matlab Use the “specgram” command; color map can be changed using “colormap gray” or similar commands 2. CSLU Toolkit Download from http: //www. cslu. ogi. edu/toolkit Registration required but free for educational use Plot spectrograms with “Speech. View” tool. 3. Praat Download from http: //www. fon. hum. uva. nl/praat/ Free and available for windows, lunix, Macintosh, etc.
Two Vowels: “preempt”
Two Vowels: “heavy oak”
Two Vowels: “reapply”
Acoustic-Phonetic Features: Manner of Articulation Approximately 8 manners of articulation: Name Vowel Approximant Nasal Stop Fricative Affricate Aspiration Flap Sub-Types vowel, diphthong liquid, glide unvoiced, voiced Examples. aa, iy, uw, eh, ow, … l, r, w, y m, n, ng p, t, k, b, d, g f, th, s, sh, v, dh, z, zh ch, jh h dx, nx Change in manner of articulation usually abrupt and visible; manner provides much information about location of phonemes.
Acoustic-Phonetic Features: Manner of Articulation Approximants (/l/, /r/, /w/, /y/): • vowel-like properties, but more constriction • /l/ has tongue-tip touching alveolar ridge, /r/ has tongue tip curled up/back (retroflex), raised and “bunched” dorsum, sides of tongue touching molars, /w/ has tongue back and lips rounded, /y/ has tongue toward front and very high • glides (/w/, /y/) can be viewed as “extreme” production of a vowel (sometimes called semivowels): /w/ /uw/ /y/ /iy/
Acoustic-Phonetic Features: Manner of Articulation Approximants (/l/, /r/, /w/, /y/): • movement of tongue slower than other vowel-to-vowel or consonant-to-vowel transitions, but not as slow as diphthong movement • sometimes voiceless when following a voiceless plosive (“play”) • /l/ may have slight discontinuity when tongue makes/breaks contact with alveolar ridge; other approximants have no discontinuity
Acoustic-Phonetic Features: Manner of Articulation Nasal (/m/, /ng/): • produced with velic port open and lips closed • sound travels through nasal cavities • these cavities filter speech with both poles (resonances) and zeros (anti-resonances) • longer pathway causes primary resonance to be low (220 -300 Hz) • anti-resonances cause higher frequencies to have lower power F 1 P 1 F 2 F 3 P 2 F 4 /m/ F 5 F 6 Z 1 Z 2
Acoustic-Phonetic Features: Manner of Articulation Nasal (/m/, /ng/): • formant structure obscured by pole-zero pairs • all three English nasals look and sound similar (place of articulation has little effect on spectrum); can be distinguished primarily by coarticulatory effects on adjacent vowel(s). • sometimes very brief duration (“camp”, “winner”) • occasional confusion with /w/, /l/ (if F 3 not visible), and closure portion of voiced stops • often sharp discontinuity with adjacent vowel • adjacent vowel may be nasalized
Acoustic-Phonetic Features: Manner of Articulation Stop (Plosive) (/p/, /t/, /k/, /b/, /d/, /g/): 1. closure along vocal tract (lips, alveolar ridge, velum) 2. buildup of air pressure behind closure 3. release of closure 4. burst of air 5. possible aspiration following burst • complex process, several changes over brief time span • some context-dependent attributes, some semi-invariant ones • voiced bursts sometimes have “voice bar” in lowfrequency region, caused by vocal fold vibration with complete oral and velic closure. • sometimes voice bar is excellent cue; sometimes can be confused with a nasal
Acoustic-Phonetic Features: Manner of Articulation /p ah p/ /t ah t/ /k ah k/
Acoustic-Phonetic Features: Manner of Articulation Stop (Plosive) (/p/, /t/, /k/, /b/, /d/, /g/): • closure and time required to build pressure results in “silence” region of spectrum prior to burst • burst airflow is a step function, which becomes similar to an impulse, which has equal energy at all frequencies • identity of a plosive contained in (at least) three areas: (1) voice-onset-time (VOT) / duration of aspiration (2) formant transitions in neighboring vowels/approximants (3) spectral shape of burst • “voiced” plosives may not show any real voicing (!)
Acoustic-Phonetic Features: Manner of Articulation Fricative (/f/, /th/, /sh/, /v/, /dh/, /zh/): • fricatives produced by forcing air through a constriction in the mouth • constriction located anywhere from the labiodental region (/f/, /v/) to palato-alveolar region (/sh/, /zh/) • all English fricatives come in voiced and unvoiced varieties • voicing may not be present in voiced fricatives (!), making duration an important distinguishing cue (voiced shorter) • the location and type of the constriction create spectral anti-resonances as well as resonances • the main difference between /s/ and /f/ is in frequencies above 4000 Hz; telephone-band speech has limit of 4 KHz.
Acoustic-Phonetic Features: Manner of Articulation Affricate (/ch/, /jh/): • Affricates are conceptually like diphthongs: two separate phonemes considered as one • English has two affricates: /ch/ /t sh/ /jh/ /d zh/ • Sometimes cue to affricate is in burst preceding fricative; in closure between vowel and fricative. • Sometimes cue to affricate is in voicing or duration.
Acoustic-Phonetic Features: Manner of Articulation Aspiration (/h/): • like vowels, except usually no voicing • can usually see formant structure • formant patterns similar to surrounding vowel(s) /ah h aw s/ = “a house”
Acoustic-Phonetic Features: Manner of Articulation Flaps (/dx/, /nx/): • allophone of /t/, /d/, or /n/ • very brief duration; no closure for /dx/ • indicated by dip in energy and F 2 near 1800 Hz “write another”
Spectrogram Reading: Fricatives • usually can divide fricatives into “strong” and “weak”: strong = /s/, /sh/, /zh/ weak = /f/, /v/, /th/, /dh/ • voicing may be present only in transition into a voiced fricative; sometimes not at all • voiced fricatives tend to be shorter than unvoiced, relative to the duration of the neighboring vowel • place of articulation causes some change in spectral shape: /sh/ and /zh/ have greater energy at lower frequency than /s/, /z/
Spectrogram Reading: Fricatives • /th/ sometimes has adjacent vowel’s F 3, F 4, F 5 extend into /th/, in contrast with /f/ • /th/ and /f/ often have weak energy during middle part of fricative • sometimes /f/ and /th/ best distinguished by formant transitions of neighboring vowel(s)… more labial vs. more alveolar characteristics of transitions. • sometimes /f/ has strong low-frequency energy (breath noise in a close-talking microphone) • sometimes /th/ has more high-frequency energy above 4 k. Hz
Spectrogram Reading: Fricatives /f iy th iy sh iy/
Spectrogram Reading: Fricatives /v iy dh iy zh iy/
Spectrogram Reading: Fricatives /f ah th ah sh ah/
Spectrogram Reading: Fricatives /v ah dh ah zh ah/
Spectrogram Reading: Fricatives /f aa th aa sh aa/
Spectrogram Reading: Fricatives /v aa dh aa zh aa/
Spectrogram Reading: Fricatives /iy f iy th iy sh/
Spectrogram Reading: Fricatives /iy v iy dh iy zh/
Spectrogram Reading: Fricatives /ah f ah th ah sh/
Spectrogram Reading: Fricatives /ah v ah dh ah zh/
Spectrogram Reading: Fricatives /aa f aa th aa sh/
Spectrogram Reading: Fricatives /aa v aa dh aa zh/