SPEAKER PREDICTING DEVICE AND SPEAKER PREDICTING METHOD

摘要

PROBLEM TO BE SOLVED: To provide a speaker predicting device capable of predicting a speaker before the speaker actually speaks.;SOLUTION: The speaker predicting device which predicts a speaker among a plurality of persons has a learning unit 1123 which learns features of an image generated prior to utterance from features of images obtained by imaging the persons, and a prediction unit 1124 which predicts the speaker to start speaking among the persons from the features of the images obtained by imaging the persons by using a learning result of the learning unit 1123. The learning section 1123 obtains the possibility that an imaged person speaks when the features of the image are generated from the features of the images obtained by the persons and uses the probability to learn features of the image generated prior to utterance. In this invention, the features of the image generated prior to the utterance are learnt to predict the speaker based upon learnt previous instances before the speaker actually speaks.;COPYRIGHT: (C)2007,JPO&INPIT