LING124 Overview of texttospeech synthesis

1
LING124 Overview of text-to-speech synthesis

• November 6, 2008

2
Text-to-speech synthesis

• Take raw text or tagged text as input and convert
  it to a speech signal
• The process is often decomposed into two parts
• Front-end for linguistic analysis of text input
• Back-end for wave-form generation

3
Front-end

• Text analysis
• Document structure detection
• Tokenization and normalization
• Linguistic analysis
• Phonetic analysis
• Grapheme-to-phone conversion
• Prosodic analysis
• Phrasing, pitch and duration attachment

4
Back-end

• Speech synthesis (wave-form generation)
• Synthesis based on vocal tract models
• Formant synthesis
• Linear prediction synthesis
• Articulatory synthesis
• http//www.haskins.yale.edu/facilities/casy.html
• Concatenative synthesis
• Di-phone
• Unit selection
• http//www.cstr.ed.ac.uk/projects/festival/