Skip to document
Welcome to StudocuSign in to access the best study resources
Guest user
0followers
Upload
HomeMy LibraryAsk AI
  • You don't have any recent items yet.
My Library
  • You don't have any courses yet.
  • You don't have any books yet.
  • You don't have any Studylists yet.

Sound Spectrography Lecture-9

Lecture slides on sound spectrography
Course

Speech and Hearing Science (SPV 349)

11 Documents
Students shared 11 documents in this course
University

Lehman College

Academic year: 2021/2022
Uploaded by:
Anonymous Student
This document has been uploaded by a student, just like you, who decided to remain anonymous.
Lehman College

Comments

Please sign in or register to post comments.

Students also viewed

Related documents

Preview text

Sound Spectrography

SOUND SPECTROGRAPHY Speech consists of a series of very complex acoustic events. The speech clinician is routinely faced with the task of evaluating speech – that is – determining in what ways speech is atypical. Because the speech signal is so complex, such an evaluation is no easy task. The various components of the acoustic signal must be separated. For clinical purposes, sound spectrography is a good method for separating the various acoustic and temporal parameters of speech. Sound spectrography is a good method because it is a complex system – one that provides different kinds of information simultaneously. A spectrographic analysis will yield a 3-dimensional output of the speech signal. It provides information about three important speech parameters:

TIME (displayed on the X-axis) FREQUENCY (displayed on the Y-axis) INTENSITY (represented by varying degrees of darkness of the markings on the spectrogram) Spectrographic analysis was traditionally performed using an instrument known as a SOUND SPECTROGRAPH. The hard-copy output of the sound spectrograph is known as SOUND SPECTROGRAM. Today, spectrography is accomplished using software associated with acoustic instrumentation such as the VisiPitch, Computerized Speech Lab (CSL) or free software such as PRAAT. Why bother with this complex speech analysis? Because sound spectrography allows the clinician to see how things change over time. What things? Fo of vocal fold vibration; vocal tract articulators If the sound spectrogram shows a change in acoustic output – some change in vocal tract configuration or vocal fold activity has occurred. e. – If formant frequencies change – we know that some change in the vocal tract transfer function has taken place (cross-sectional area of the vocal tract has changed). Recall – the formant frequencies seen on a wide band spectrogram represent the resonant frequencies of the vocal tract. You can’t draw a direct link between the acoustic output on a spectrogram and the physiologic adjustments of the vocal tract. Speech is just too complex. However, you can draw meaningful inferences about vocal tract dynamics from acoustic data. Thus – you can pinpoint problems which can then be tested more directly (Example – perception of instability during phonation. Source of instability? Vocal tract, vocal folds, or both?)

Sound spectrography is a non-invasive method (the signal is captured using a

Notes

from same

lesson /

####### corresponds to

PP →

  • Fo = fundamental

####### frequency

was higher during production of this vowel. Why is fo changing – fo changes with stress and intonation patterns. In the last two sound segments of this word (/ən/) there is a falling intonation contour – how do we know that? Striations get farther apart during the production of /ən/. Important point: you can get information about suprasegmentals from a spectrogram!

Look at the formant structure during the last /ə/ production. Formants are relatively stable, even though fo is not. Significance? To a large degree fo and formant frequency changes are independent of one another.

What does a stable formant pattern reveal? That the vocal tract resonator was held in a stable configuration. What do evenly-spaced vertical striations suggest? Stable fo of vocal fold vibration. SLIDE; JOE TOOK

FATHER’S SHOE BENCH OUT

Sound spectrography provides an excellent overall picture of a number of features of speech. It permits us to see how the vocal folds and vocal tract adjust over time. We can see when, in what order, and with what stability and timing – things change. SLIDE: HARMONIC INSTABILITY

This is a classic NARROW BAND spectrogram (generated using a 45 – 50 Hz bandwidth analyzing filter) Here we see shifting harmonics – this means there has been a change at the glottal source (fo has shifted) If the articulators in the supra-laryngeal vocal tract change on what spectrogram would you expect to see this?

On a wide-band spectrogram – shifting formants The SLP can document instability in the vocal tract, at the level of the larynx, or both simultaneously in a matter of minutes by evaluating a patient’s production of a sustained vowel. Sound spectrography can answer the question: WHERE IN THE PERIPHERAL SPEECH SYSTEM DOES MY PERCEPTION OF INSTABILITY ORIGINATE? THE LARYNX, THE VOCAL TRACT, BOTH??? Knowing the answer to this question can prevent malpractice – e. – working on the larynx when the problem really exists in the supralaryngeal resonator. Making a spectrogram is an art – interpreting a spectrogram is a science which requires knowledge of vocal tract structure and function.

Wide Band Spectrogram

  • A wide band spectrogram of

“We were away a year ago”, which is almost entirely voiced. Notice, due to the improved temporal resolution, the vertical striations representing individual glottal pulses. Also evident are formant frequency bands. Note the shifting formant bands reflecting the articulatory adjustments of the vocal tract (changes in cross- sectional area)

Vowel Formant Structure

  • Each vowel is created by a distinct vocal

tract configuration which results in a unique

resonant frequency pattern.

  • The resonant frequencies are shown on a

wide band spectrogram as formants.

  • Formant structure for /i/, /I/, /e/, /ε/, and /ǽ/

Formant Frequency Measurement

  • Linear Predictive Coding (LPC) analysis is used to track the center

frequencies of formant bands

  • Note the shifting formants

(particularly F1 and F2) as vocal tract cross sectional changes during diphthong

production

Wide (Broad) Band Spectrogram:
Vowels /u/ and /a/

Sound Pressure Waveform is shown above wide band spectrogram

####### Spectrograms: Visualizing the Spectral

####### Features of Speech

  • Monophthong vowel /a/: Formants are shown as horizontal bands of high energy (red- green/yellow). Blue regions illustrate decreasing energy. Note the formants are stable over time demonstrating an unchanging SLVT cross-sectional area. F1 and F3 are shown (F2 has been absorbed into F3)

  • Diphthong /aІ/: note longer duration than monopthong and shifting formant structure as cross- sectional area of the vocal tracts changes

Wide Band Spectrograms
Vowels F1 & F2 Locations &
Transitions

bV

dV

gV

Acoustic cues for manner of
articulation in consonants:
  • Speech can be roughly segmented into manner of

articulation categories by context-free acoustic cues.

  • Stops: Closure (silent) period followed by release burst

and abrupt vowel onset.

  • Fricatives: Turbulent noise burst, strong for sibilants,

weak for non-sibilants.

  • Nasals: Abrupt onset and offset of a segment with very

weak formant structure. Low frequency, periodic energy (voice bar on spectrogram).

  • Approximants: Non-abrupt onset and offset; dynamic

(changing) formant structure (diphthong-like); weaker F and F3 than for (more open) vowels.

Fricatives and Stop-Plosives

  • Top figure: /sev/

  • Lower figure: /taΙd/

  • Note relatively longer duration

of /s/ vs. /t/ (continuous vs. transient sounds)

  • /s/ is characterized by high

frequency noise; /v/, which is combined noise and quasi- periodic vibration, shows evidence of both. Note low

frequency voice bar.

  • /t/, a transient sound is shorter

in duration than the /s/; the syllable arresting /d/ was released

Non-sibilant fricatives

Kent & Read, (2002) p-

  • Wide spectral energy distribution
  • Note the effect of voicing on the fricative spectrum.
  • The presence of a low frequency ‘voice bar’.

Wide Band Spectrogram: /si/

Frequency range: DC – 5400 Hz

Was this document helpful?

Sound Spectrography Lecture-9

Course: Speech and Hearing Science (SPV 349)

11 Documents
Students shared 11 documents in this course

University: Lehman College

Was this document helpful?
Sound Spectrography

Students also viewed

Related documents