| 摘要 |
PURPOSE: To provide a speech recognition device capable of easily revising and adding a vocabulary, obtaining high recognition performance and easily dealing with free talking also. CONSTITUTION: This device is provided with a speech analysis part 101 extracting characteristic parameter of an input voice, a collation part 104 collating the characteristic parameter with a word standard pattern, the collating part 110 collating the characteristic parameter with the word standard pattern formed from a vocal sound standard pattern, a likeness decision part 106 and a likelihood decision part 113 respectively extracting word candidate systems of a first stage from the collation result of the collation parts 104 and 110, construction/meaning analysis parts 107 and 114 performing construction/meaning analysis for the word candidate systems of the first stage and outputting the word candidate systems of a second stage, a synthetic decision part 109 synthetically deciding the word candidate systems of the second stage and outputting the word candidate systems of a final stage and the construction/meaning analysis part 115 performing the construction/meaning analysis for the word candidate systems of the final stage and obtaining a speech recognition result. |
