| 摘要 |
PROBLEM TO BE SOLVED: To specify and to take out a segment, which includes key information in a voice message, by putting a tag on the voice message to indicate the position of the key segment detected in the message and taking out the key segment from the message. SOLUTION: A key segment, that is to be registered, is inputted by marking the segment in a message. The segment is processed by using a voice recognition 12 and pronunciation of the segment is generated. Then, an identifier of the key segment and the feature corresponding to the segment are stored in a segment name and feature (pronunciation) 15. Detection of the key segment is executed by using independent models for the feature of the key segment stored in the segment name and feature (pronunciation) 15 and a speaker and a spot is put on the key segment registered in the voice message. The detection of a key segment executed in the key segment detection is executed by a conventional word spot. |
