PRODUCTS
|
|
 PRESS RELEASE
May 8, 2009
GritTec laboratory announced a new version of automatic text independent speaker identification technology.
The technology of automatic text independent speaker identification is
intended for identification in non real-time mode of an unknown
audio record by pair comparison with samples of audio records of
target speakers.
Changes have concerned the module of calculation of probabilities
observation and modules of building and estimation of 1-th (FRR)
and 2-th (FAR) sort errors. In particular in the algorithmic part re-estimation of probabilities observation has been
changed on the basis of maximization of probability observation on S-states. In the module
of errors building (FRR, FAR) the technology of FAR
construction and its estimation in identification mode has been changed.
Structures used at the description of sample audio record of target speaker have been changed:
- structure of voice model;
- structure of errors model (FAR, FRR, ERR);
- structure of speaker model consisting from the structures of voice model,
errors model and model of background noise and channel distortion.
So, in the structure of voice model it has been optimized the
statistical description of Gaussian mixture models (GMM) taking into
account S-states and probabilities of transition between them.
GritTec Ltd. will advance technology of voice identification in the
market of Voice Biometric under mark of GritTec’s Speaker-ID (see urls:
www.speaker-id.com,
www.speaker-id.ru) in the
form of the commercial version of identification system for small
databases (no more than 5 - 10 thousand samples of audio
records of target speakers) and for developers in the package of SDK
libraries.
At the current moment GritTec’s Speaker-ID has passed preliminary
tests in office conditions and on the specialized audio databases of
real telephone records. The package of SDK libraries of new version GritTec’s Speaker-ID and its demo version will be available for developers in the near future.
About Grittec
GritTec Laboratory (GritTec Ltd.) specializes on research and
development of algorithms and technologies in the field of speech
and audio processing. GritTec's research is focused on speech
enhancement, speech concealment, voice biometric, speech recognition,
speech synthesis and other speech and audio technologies.
Url: http://www.grittec.com
|
|
|
|
