Speech recognition apparatus for utterance speech of foreigners

摘要

A speech recognition apparatus according to the present invention includes a feature extraction unit for extracting information useful for recognition from an input speech and converting the extracted information into a feature vector; An acoustic model database for storing a predetermined acoustic model; A language model database for storing a predetermined language model; A pronunciation model database for storing a pronunciation model further including a variation sound generated by a grammatical error that can be committed by a foreign person in a pronunciation model of a native speaker; A search unit for searching a word sequence having the highest probability using the acoustic model database, the pronunciation model database, and the language model database based on the feature vector; And a recognition result output unit for providing a recognition result of the input speech using the output of the search unit.