Extended recognition dictionary study device and speech recognition system

摘要

Vis-a-vis the speaker who is used anew, using the utterance label which corresponds with the sound of that speaker in advance, the necessary dying which you study, it was suited for that speaker, recognition making use of the extended recognition dictionary is made possible.The extended recognition dictionary study device comparing with the acoustic model series which is output from the speech recognition result and the correct answer acoustic model series which is input, the utterance deformation which appears widely correspondence of those models as the utterance deformation data in the utterance deformation data calculation expedient which it calculates and the utterance deformation data which was calculated and being inclined, makes plural utterance deformation gathering combining the utterance deformation data taxonomic expedient which classifies the utterance deformation which appears and the utterance deformation which is classified, possesses with the recognition dictionary extended expedient which expands the recognition dictionary in every utterance deformation gathering with the utterance deformation which is included in the respective utterance deformation gathering,. The speech input system outputs the speech recognition result making use of the extended recognition dictionary every of this utterance deformation gathering.