| 摘要 |
<p>In order to provide a speech processor that detects points of change in speech characteristics from an audio signal in audio-visual content, this speech processor is provided with: a characteristic calculation means for calculating segment characteristics that express the characteristics of an inputted audio signal for each unit segment having a prescribed time length in the audio signal; a boundary calculation means for calculating boundary information related to the boundaries between similar segments, which comprise a plurality of the unit segments with similar segment characteristics, such as the unit segments, and other segments, with respect to each of the plurality of unit segments; and a determination means for calculating the priority of each boundary represented by the boundary information, and determining whether or not each of the boundaries represents a scene change point, on the basis of the priority.</p> |
