Skip to content

Speech Recognition Engines

douglasbagnall edited this page Jul 24, 2012 · 1 revision

CMU Sphinx is a family, not a single decoder, though all are kept in the same subversion tree and share a 4 clause BSD licence.

  • Written in C
  • tries to be fast
  • There are many users and existing models
  • Java
  • aimed more at flexibility and research than speed

“HMM Tool Kit”.

  • Silly license.

HTK is a lot more than a speech decoder, and most other engines depend on it for generating models.

Uses models derived from HTK.

  • Custom sort of BSD-ish licence.

Few users and developers?

Few users and developers?