| 摘要 |
PROBLEM TO BE SOLVED: To simply and speedily extract features. SOLUTION: A time-interval analyzing part 81 temporally and continuously analyzes temporally continuous voice data in each prescribed frequency band. A continuous music feature extraction part 82 extracts a continuous feature being a temporally continuous feature from the analyzed results. A cutout part 83 cuts out a part of a period of a preset length from the continuous feature. An interval feature extraction part 84 extracts an interval feature being the feature expressed by any one of one scalar or vector respectively in each of periods in each period from the part of a cut-out continuous feature. A music/conversation discrimination part 85 estimates a target feature being the feature indicating one feature of the voice data of an entire period in each period from the interval feature. The invention can be applied to a recording and reproducing device for recording and reproducing the voice data. COPYRIGHT: (C)2008,JPO&INPIT
|
