摘要 |
PURPOSE: A method for constructing a voice database is provided to recognize a voice by constructing a voice database in phoneme like units. CONSTITUTION: The method for constructing a voice database comprises the following steps of a collection process(S10), an extraction process(S20), and a labelling process(S30). The collection process(S10) is to collect a word or a sentence from a particular speaker. The extraction process(S20) is to extract phoneme like units predicted from the collected voice of the specific speaker. The labelling process(S30) is to designate specific extension names and file names of each extracted phoneme like unit and store them to a voice database.
|