Method and system adaptive to speaker without vetating to speech identifying database
This invention discloses a method and a system for generating phonetic identification acoustic formwork data base. Said method includes: receiving trained phone wave form from a first speaker to be divided into blocks corresponding to HMM states, computing Gauss PDF related to the speaker to every H...
Saved in:
Main Author | |
---|---|
Format | Patent |
Language | English |
Published |
15.06.2005
|
Edition | 7 |
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | This invention discloses a method and a system for generating phonetic identification acoustic formwork data base. Said method includes: receiving trained phone wave form from a first speaker to be divided into blocks corresponding to HMM states, computing Gauss PDF related to the speaker to every HMM state. In the acoustic formwork database not related to speakers, the speaker-related PDF replaces that not related, among which, the not-related database includes multiple speaker-not related HMM, each one contains multiple Gauss PDF, each has original weight coefficient, new weight coefficient is computed for each related or not related PDF in the not related database. |
---|---|
Bibliography: | Application Number: CN20031118283 |