Title:
UTTERANCE SEGMENT DETECTION DEVICE, UTTERANCE SEGMENT DETECTION METHOD, AND PROGRAM
Document Type and Number:
WIPO Patent Application WO/2021/014612
Kind Code:
A1
Abstract:
Provided is an utterance segment detection device with which it is possible to detect an utterance segment with high precision on the basis of whether the end of a voice segment is an utterance end. The present invention includes: a voice/non-voice assessment unit that performs a voice/non-voice assessment, which is an assessment as to whether a given frame of an acoustic signal is a voice; an utterance end assessment unit that performs, for each voice segment, an utterance end assessment, which is an assessment as to whether the end of the segment is an utterance end, the voice segments being segments for which the result of the voice/non-voice assessment is that the segment includes a voice; a non-voice-segment continuation time threshold value determination unit that determines a threshold value pertaining to the continuation time of a non-voice segment on the basis of the result of the utterance end assessment; and an utterance segment detection unit that compares the continuation time of a non-voice segment following a voice segment with the corresponding threshold value, and detects an utterance segment.
Inventors:
MASUMURA RYO (JP)
OBA TAKANOBU (JP)
MATSUI KIYOAKI (JP)
OBA TAKANOBU (JP)
MATSUI KIYOAKI (JP)
Application Number:
PCT/JP2019/029035
Publication Date:
January 28, 2021
Filing Date:
July 24, 2019
Export Citation:
Assignee:
NIPPON TELEGRAPH & TELEPHONE (JP)
International Classes:
G10L15/04; G10L25/78
Foreign References:
| JP2005017932A | 2005-01-20 | |||
| JP2019040148A | 2019-03-14 | |||
| JPH01219893A | 1989-09-01 | |||
| JP2017078848A | 2017-04-27 |
Attorney, Agent or Firm:
NAKAO, Naoki et al. (JP)
Download PDF: