PERSONAL NAME ASSIGNMENT APPARATUS AND METHOD

摘要

PROBLEM TO BE SOLVED: To specify a scene where a desired performer appears based only on a received video picture.

SOLUTION: The personal name assignment apparatus includes: a means 103 for obtaining a speaker time period as a first section including a speaker name specified by information indicating the speaker name and an utterance time period of the speaker; a means 101 for obtaining a second section including an utterance from a non-silent section in the video picture; a means 105 for extracting a first feature amount characterizing the speaker from a speech waveform of the second section when the second section is included in the first section and associating the speaker name corresponding to the first section with the feature amount; a means 106 for creating a speaker model of the speaker from the feature amount for each speaker; a means 108 for obtaining a third section that is an utterance time period to be recognized; a means 109 for extracting a second feature amount characterizing the speaker from a speech waveform of the second section when the second section is included in the third section; a means 110 for calculating similarity between the feature amount of the speaker model for each speaker and the second feature amount; and a means 111 for recognizing the speaker name of the speaker model of a setting condition as the performer, in the similarity.

COPYRIGHT: (C)2010,JPO&INPIT