Effective Open Source Speech Recognition in Your Application - PowerPoint PPT Presentation

Effective Open Source Speech Recognition in Your Application #kde-speech Peter Grasch peter@grasch.net

The Basics Speech model Decoder Acoustic model Language model ● Sounds ● Vocabulary ● Grammar

Open Source Speech Recognition Decoder Trainer UI CMU SPHINX ✓ ✓ (PocketSphinx, SphinxTrain) Julius ✓ KALDI ✓ ✓ Simon ✓ ✓ ✓

Standard Architecture Commands Simond Simon Your application ? Acoustic model Language model

Standard Architecture Commands Simond Simon Scenario Scenario Your application Scenario Acoustic model Language model

Headless Architecture Commands Simond Simon Your application Acoustic model Language model

Embedded Architecture Commands Simond Simon Your application Acoustic model Language model Decoder

Standard Architecture Commands Simond Simon Scenario Scenario Your application Scenario Acoustic model Language model

Writing your Scenario ● Lay out the commands you want to support ● Create: – Vocabulary – Grammar – Commands

Writing your Scenario Demonstration

Tighter Integration: Write a Custom Command Plug-In ● Full, programmatic control of the scenario ● Meta information of recognition results: – Phonetic transcriptions – Confidence scores* – Alternative results*

Tighter Integration: Write a Custom Command Plug-In Demonstration

Q & A #kde-speech Peter Grasch peter@grasch.net

Thank you for your attention