VoxForge collects transcribed user-submitted speech audio files (collectively called a "speech corpus") to create acoustic models for use with speech recognition engines such as HTK, Julius, ISIP, and Sphinx. The current focus is on collecting audio to create acoustic models for command and control applications on a PC, and for voice over IP telephony speech recognition applications, i.e. IVR (Interactive Voice Response).
|Tags||Communications Telephony Desktop Environment multimedia Sound/Audio Speech|
|Operating Systems||POSIX Linux Windows Cygwin|
Release Notes: New audio files used were used to create an updated acoustic model.
Release Notes: The QuickStart Acoustic Model was updated.
Release Notes: This QuickStart download is designed to highlight the use of VoxForge acoustic models with Open Source speech recognition engines, starting with the Julius engine. The acoustic models currently included are still in alpha stage, so the recognition quality reflects this. Much more GPL-transcribed speech audio is needed to create decent-quality acoustic models; users are encouraged to submit transcribed speech to VoxForge.
No changes have been submitted for this release.