| 摘要 |
PURPOSE:To prevent the generation of misdetection due to noise, to specify a real speaking voice section, to specify a speaker corresponding to voice detection or speaking importance, and to smoothly steering a conference by detecting only a voice out of plural noises in the conference by means of a voice judging part. CONSTITUTION:A voice judging part 1 judges and sorts input voice data sampled by N voice detection parts 3 in each fixed period to a voice section and a non-voice section and a voiceless sound or a noise is judged from a voice sound. A channel in speaking in a conference is detected as a speaker selecting candidate based upon a fixed period in which voice judging results for N channels which are outputted from the voice detection parts 3 corresponding to respective microphones in an integrated judging part 1 and a threshold. A speaker determining part 8 determines a speaker based upon the conditions of a priority setting part 6 and a comparing part 7 and stores a voice detecting section. The generation of detection error due to a noise can be prevented by allowing the voice judging part 1 to detect only a voice from the noise. |
