Speech Production - PowerPoint PPT Presentation

1 / 21
About This Presentation
Title:

Speech Production

Description:

Explain how we produce speech sounds. Explain the concepts of phonemes, voiced and unvoiced sound, vowels and formants ... (c/w 10-20 dB in cochlea implant) ... – PowerPoint PPT presentation

Number of Views:642
Avg rating:5.0/5.0
Slides: 22
Provided by: billd59
Category:

less

Transcript and Presenter's Notes

Title: Speech Production


1
Speech Production
Introduction to Acoustics
  • University of Salford
  • Acoustics Audio and Video Group

2
Learning Outcomes
  • Explain how we produce speech sounds
  • Explain the concepts of phonemes, voiced and
    unvoiced sound, vowels and formants
  • Discuss a simple acoustic model of the voice

3
Basic questions
  • How do we make sound with the voice?
  • How do we control this sound to talk?

4
Organs of speech
5
Speech sounds
  • Phonation vocal folds vibrate periodically when
    air forced through them e.g. vowels
  • Fricatives air forced through constriction in
    vocal tract e.g. s, f
  • Plosives sharp close then open of vocal tract
    e.g. p
  • Phonation is voiced, others are unvoiced

6
Vocal folds open closed
7
Model of phonation
8
Characteristics of phonation
9
Phonemes
  • How to describe speech sounds?
  • Is written language adequate?
  • Phonetic symbols denote actual speech sounds
    phonemes
  • Different phonemes are formed by using different
    combinations of nasal cavity, tongue and lips

10
Phonetic alphabet
  • See phoneme list, over

11
Vowel sounds exercise
  • By self-experimentation, determine the location
    of these vowel sounds on the chart opposite
  • /i/ (heed)
  • /?/ (had)
  • /u/ (food)

12
Formant frequencies
13
Formants exercise
  • There is a rough correlation between tongue
    position and the formant frequencies F1 and F2
  • Study the formant table for /i/ and /a/ and see
    if you can label your previous vowel diagram

14
Real speech time domain
15
Spectrogram example
16
Speech levels at 1 m, free field
17
Directivity
18
Dynamic range
  • Usual assumption is 30 dB
  • Experimental evidence for 50 dB or more
  • (c/w 10-20 dB in cochlea implant)
  • Normal hearing 60 correct words in sentences
    at 0 dB S/N 95 at 18 dB

19
Speech synthesis
20
Speech synthesis how?
  • NLP
  • large database of rules
  • linguistics
  • related to AI
  • DSP
  • ideal is physical model of vocal tract (like our
    one, but better)
  • practice is recorded phonemes or combinations of
    phonemes

21
Conclusions
  • Simple model of voice source (glottal pulse) -gt
    variable filter (vocal tract)
  • Phonemes are speech sounds
  • Vowel sounds are identified by formant frequencies
Write a Comment
User Comments (0)
About PowerShow.com