公开(公告)号：US11120819B2

公开(公告)日：2021-09-14

申请号：US16122338

申请日：2018-09-05

Applicant: YAHOO JAPAN CORPORATION

Inventor： Motoi Omachi

IPC: G10L15/22 ,  G10L25/18 ,  H04R5/04 ,  H04R3/00 ,  G10L25/78 ,  G10L21/0208 ,  G10L21/0216 ,  H04R1/40

Abstract: A voice extraction device according to the present invention includes a formation unit, an acquisition unit, an emphasis unit, a generation unit, and a selection unit. The formation unit forms directivity through beam-forming processing for each microphone in a microphone array including a plurality of microphones that form a plurality of channels. The acquisition unit acquires an observation signal that is a signal of voice received by each of the channels. The emphasis unit generates an emphasized signal by emphasizing the observation signal in accordance with the directivity formed by the formation unit. The generation unit generates, for each channel, frequency distribution of amplitude of the emphasized signal generated by the emphasis unit. The selection unit selects a channel corresponding to a voice signal used for voice recognition from among the channels based on the frequency distribution corresponding to the respective channels generated by the generation unit.