To precisely perform speech recognition even in various scenes such that background noise environment changes during speech input from a communication terminal.
A speech recognition system is equipped with a communication terminal 1 and a storage means 2 connected through a communication network 200. The communication terminal 1 has a speech input means 11 of inputting a speech and ambient noise, a feature extracting means 12 of extracting a feature quantity of the speech input from the speech input means 11, a speech recognition means 13 of recognizing the speech by collating the feature quantity of the speech with a specified recognition model parameter, and a noise verifying means 14 of verifying the ambient noise input from the speech input means 11. The noise verifying means 14 selects a recognition model parameter corresponding to the kind of the verified noise from a plurality of recognition model parameter sets stored in the storage means 2 when the ambient noise is verified and the speech recognition means 13 performs speech recognition by using the recognition model parameter selected by the noise verifying means 14.
COPYRIGHT: (C)2007,JPO&INPIT
JP3276196A | ||||
JP62042198A | ||||
JP1260495A | ||||
JP8006589A | ||||
JP61091700A | ||||
JP9507105A | ||||
JP10282990A | ||||
JP2000194392A |
WO1999021172A1 |