| 摘要 |
PURPOSE:To stably detect an unknown word by correcting similarity so that the similarity can be compared on the same scale regardless of differences of a speaker and environment. CONSTITUTION:A comparative collation part 2 calculates the similarity of the feature vector of an input voice to plural standard patterns which are registered in advance. At this time, a standard pattern which gives a maximum value S of similarity is found as a recognition result. Simultaneously, a reference similarity calculation part 4 compares the feature vector V with the standard pattern formed by connecting unit standard patterns in a unit standard pattern storage part 3. Here, the maximum value of similarity is outputted as reference similarity R. Then a similarity correction part 5 uses the reference similarity R to correct the similarity S. |
