| 摘要 |
<P>PROBLEM TO BE SOLVED: To discriminate a speaker simply and correctly. <P>SOLUTION: A speaker discrimination device 50 obtains respective voice data from a plurality of microphones disposed for respective speakers, and frames the obtained voice data into frames in predetermined sections, and identifies whether each frame is a voiced sound region or a voiceless sound region on the basis of a first probability model, and when identifying in duplication the voiced sound region in the frames in the same section for the respective voice data, validates the identification result of the frame having maximum energy among the frames identified as the voiced sound region in the same section, and identifies a speech region and a silent region in the respective voice data from the validated identification result of the frame on the basis of a second probability model. <P>COPYRIGHT: (C)2013,JPO&INPIT |
