Week 15. Speech Engineering Overview - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Week 15. Speech Engineering Overview

Description:

Text-to-speech (TTS) systems. Reviving Sonus. 4. Speech ... Text-to-speech conversion processes. Data input. Linguistic Process. Morphological Analysis ... – PowerPoint PPT presentation

Number of Views:21
Avg rating:3.0/5.0
Slides: 12
Provided by: taeyeo
Category:

less

Transcript and Presenter's Notes

Title: Week 15. Speech Engineering Overview


1
Week 15. Speech EngineeringOverview
  • Linguistic background
  • Phonetics
  • Phonology
  • Engineering background
  • Acoustics
  • Mathematics
  • Computer science

2
Subfields of application
  • Speech Recognition
  • Let machines hear
  • Simulating human speech perception
  • Speech Synthesis
  • Let machines say
  • Simulating human speech production
  • Automatic Translation
  • Ultimate goal
  • Needs other sub-areas

3
Speech synthesis- short history
  • Before 1970s
  • Analogue vocal tract model
  • After 1970s
  • PC hardwares
  • After 1990s
  • PC softwares
  • Text-to-speech (TTS) systems

4
Speech synthesis- basic processes
  • Recording of human voice
  • Cutting into units
  • Coping and pasting of various units
  • Units
  • phones, phonemes, syllables, words

5
Speech synthesis- problems
  • Transitions and coarticulation
  • Solution Introducing new units
  • diphones
  • triphones
  • demisyllables
  • Prosody
  • Solution
  • duration modelling
  • pitch modelling

6
Text-to-speech conversion processes
  • Data input
  • Linguistic Process
  • Morphological Analysis
  • Syntactic Analysis
  • Letter-to-Phoneme conversion
  • Prosodic control
  • Speech generation

7
Speech recognition- application
  • Voice dictation
  • Automobile maneuver control
  • Operation of home appliances
  • Speaker identification
  • Speaker verification

8
Speech recognition- short history
  • 1930s getting interested
  • 1940s spectrograph invented
  • 1950s, 60s isolated word recognition
  • 1970s interested in connected speech
  • 1980s statistical approach

9
Speech recognition-quality control
  • Word sequence
  • Isolated words / Connected speech
  • Naturality
  • Read speech / Spontaneous speech
  • Spontaneous dialogue
  • Speaker
  • Speaker dependent/independent system
  • Language
  • Language dependent/independent system
  • Vocabulary size
  • Small / medium/ large size vocabulary
  • Vocabulary independent system

10
Speech recognition- units
  • Trade-off
  • Efficiency and performance
  • Units
  • Words
  • Phone-like units
  • Diphones
  • Triphones

11
Speech recognition- Teniques
  • Knowledge-based approach
  • Statistical approach
Write a Comment
User Comments (0)
About PowerShow.com