| 摘要 |
PURPOSE:To realize the word voice recognizer which is easily adapted to a speaker and has a small storage capacity required for the adaption. CONSTITUTION:This device consists of a voice input means 1, a feature extracting means 2, a feature understanding extent vector extracting means 3, a recognizing and registering operation control means 4, a registered word feature similarity vector storage means 5, and a pattern matching means 6. An adapting process is performed only by voicing 'a', 'i', 'u', 'e', 'o', 'N', etc., by the speaker and data on registered words are usable in common among speakers. With those adaption data, a completely new word group can be recognized by a completely new device without any adapting procedure. Further, individuals can be confirmed by using them. Further, those data are extremely small in volume, so the storage capacity required to hold the data is made extremely small even on a voice recognition system for many persons. |
