Speech production

Question 1

3 sources of speech production

Answer 1

All three have wideband spectrum:

• Voicing: vibration of the vocal folds, same type of aerodynamic mechanism as a flag flapping in the wind.
• Frication or Aspiration: turbulence created when air passes through a narrow aperture
• Burst: the “pop” that occurs when high air pressure is suddenly released

Question 2

3 steps to produce speech

Answer 2

1. initiation
2. phonation
3. articulation

Question 3

vocal folds aka cords

Answer 3

two bands of smooth muscle tissue found in the larynx (voice box).

Question 4

vocal tract

Answer 4

air cavity between glottis and lips

Question 5

source-filter model of the vocal tract

Answer 5

Question 6

source-filter model of the vocal tract (with details)

Answer 6

Question 7

formula for speech signal (through convolution of excitation signal and transfer function)

Answer 7

The speech signal s(t), is created by convolving (∗) an excitation
signal e(t) through a vocal tract transfer function h(t):
s(t) = e(t) * h(t)

Question 8

Fourier transform through excitation product times transfer function

Answer 8

The Fourier transform of speech is the product of excitation
times transfer function:

S(f) = H(f)E(f)

Question 9

formants

Answer 9

vocal tract resonances

Question 10

what happens at resonant frequencies?

Answer 10

At the resonant frequencies, the resonance enhances the energy of
the excitation, so the transfer function H(f) is large at those
frequencies, and small at other frequencies

Question 11

air stream (vowels)

Answer 11

unblocked air stream

Question 12

air stream (consonants)

Answer 12

blocked / obstructed air stream

Question 13

vowels scheme

Answer 13

Question 14

vowel classification: tongue height

Answer 14

Tongue height:
– Low: e.g., /a/
– Mid: e.g., /e/
– High: e.g., /i/

Question 15

vowel classification: tongue advancement

Answer 15

Tongue advancement:
– Front : e.g., /i/
– Central : e.g., /ə/
– Back : e.g., /u/

Question 16

vowel classification: lip rounding

Answer 16

Lip rounding:
– Unrounded: e.g., /ɪ, ɛ, e, ǝ/
– Rounded: e.g., /u, o, ɔ/

Question 17

vowel classification: tense vs lax

Answer 17

Tense/lax:
– Tense: e.g., /i, e, u, o, ɔ, ɑ/
– Lax: e.g., /ɪ, ɛ, æ, ə/

Question 18

vowel scheme: dependence of formants 1 and 2 on tongue placement

Answer 18

Question 19

consonants classification: manners of articulation

Answer 19

Manner of articulation
– Stops: /p, t, k, b, d, g/
– Fricatives: /f, s, S, v, z, Z/
– Affricates: /tS, dZ/
– Approximants/Liquids: /l, r, w, j/
– Nasals: /m, n, ng/

Question 20

coarticulation

Answer 20

• Coarticulation refers to changes in speech articulation (acoustic or visual) of the current speech segment (phoneme or viseme) due to neighboring speech. In the visual domain, this phenomenon arises because the visual articulator movements are affected by the neighboring visemes.
• production of a speech sound becomes
  more like that of a preceding/following speech sound

Question 21

f0 and H1

Answer 21

• fundamental frequency f(0) and first harmonic H(1) are the same thing.
• The fundamental frequency, or f0, is the first harmonic, or H1. There is a harmonic at each interval of the f0 up to infinity. Vocal fold vibration produces many harmonics above f0, all the way up to 5000Hz in the adult human vocal tract. These harmonics decrease in amplitude as the frequency increases.

Question 22

magnitude spectrum and log magnitude spectrum formulas

Answer 22

magnitude spectrum: S(f) = H(f)E(f)

log magnitude spectrum = ln |S(f)| = ln |H(f)| + ln |E(f)|

Question 23

axes(spectrogram)

Answer 23

Spectrogram = time on the horizontal axis, frequency on vertical axis.

Question 24

spectral splatter

Answer 24

spectral splatter (also called switch noise) refers to spurious emissions that result from an abrupt change in the transmitted signal, usually when transmission is started or stopped.

For example, a device transmitting a sine wave produces a single peak in the frequency spectrum; however, if the device abruptly starts or stops transmitting this sine wave, it will emit noise at frequencies other than the frequency of the sine wave. This noise is known as spectral splatter.

When the signal is represented in the time domain, an abrupt change may not be visually apparent; in the frequency domain, however, the abrupt change causes the appearance of spikes at various frequencies.