| 摘要 |
PROBLEM TO BE SOLVED: To provide a voice information extracting device capable of accurately extracting a human voice part from content even if the content contains noise, background music, or the like. SOLUTION: A voice feature extracting part 2 analyzes input voice and outputs a voice feature parameter. A collation part 3 collates the voice feature parameter extracted by the voice feature extracting part 2 with a human voice detecting standard pattern 5, and outputs a result of the collation to a human voice extracting part 1 and a human voice extraction point deciding part 4. When non-human voice section continues, the human voice extraction point deciding part 4 outputs a detection instruction to the human voice extracting part 1 to extract a human voice section. The human voice extracting part 1 starts human voice extraction processing according to the detection instruction from the human voice extraction point decision part 4. When the collation result from a collation part 3 is a human voice section, the voice of the corresponding section is extracted through the human voice extraction processing. COPYRIGHT: (C)2006,JPO&NCIPI
|
