To provide a method and a device for voice recognition which can improve a recognition rate by being used for a communication service system capable of receiving an outgoing ID at the time of termination and a medium where their program is recorded.
An outgoing ID decision means 5 checks whether or not the outgoing ID at the time of termination is a caller's number and when the outgoing ID is not a caller's telephone number, a parameter set for general telephone is loaded from a 1st storage means 2 to a voice recognition engine 1, but when the outgoing ID is a caller's number, an acoustic environment specifying means 6 specifies acoustic environment and a parameter set corresponding to it is loaded from the 1st storage means 2 to the voice recognition engine 1. Further, when a caller retrieving means 7 can retrieve a speaker adaptation parameter corresponding to the specified acoustic environment and corresponding to the caller's telephone number from a 2nd storage means 3, an arithmetic means 4 is used to optimize the parameter set loaded to the voice recognition engine 1.
NARAHARA YOSHIORI
HOSOBUCHI TAKASHI
Next Patent: VOICE RECOGNITION DEVICE