ISRSpeaker Interface

The ISRSpeaker interface allows the transfer of speaker information, or the speaker profile, from one system to another, so that a speaker doesn't have to perform training again when he or she switches to another system. For command and control applications, the effort of transferring a speaker profile may outweigh the benefits; however, it is extremely useful for dictation systems (which often require a lot of training) and speech servers (which need to share training). An engine may or may not support this interface.

The speaker profile should include all of the information that the engine has about the speaker. For example, a speaker profile can include:

· A data header to make sure that the application isn't passing information to the wrong engine

· The languages for which training has been done

· Known patterns of speech and the language model

· How specific words are pronounced

· Phonetic training

· Speaker identity

· Speaker preferences

· The number of hours of training done to create the profile (useful for merging speakers)

The ISRSpeaker interface is supported by the speech-recognition engine object. It has these member functions: