Selective speaker conformity method for load speech recognition system

摘要

(57) Abstract The method of improving the recognition precision of the load speech recognition system is disclosed. Method of this invention, using N best conformity technology, makes the audio engine of the system the audio quality of the speaker conform. In this method, as for the speech recognition system, receiving the speech which possesses the audio quality of the speaker of specification, pertaining to automobile order to process (76), it selects the N best audio instruction set which conforms to speech (78). When training mode input from the speaker is received, as for the system, to output N best instruction set (86), as for the speaker accurate order is selected in the speaker. Next, the system in order the speech which possesses the audio quality which is received to be recognized as the automobile order which is selected by the user, makes the audio engine conform (90).