Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD AND COMPUTER PROGRAM FOR GENERATING VOICE FOR EACH INDIVIDUAL SPEAKER
Document Type and Number:
WIPO Patent Application WO/2021/002649
Kind Code:
A1
Abstract:
According to an embodiment of the present invention, a method for generating a voice for each individual speaker from a voice content including a section in which at least two speakers speak concurrently may comprise the steps of: dividing the voice content into one or more single-speaker sections and one or more multi-speaker sections; determining speaker feature values corresponding to the one or more single-speaker sections, respectively; generating grouping information by grouping the one or more single-speaker sections on the basis of similarity of the determined speaker feature values; determining a speaker feature value for each individual speaker by referring to the grouping information; and generating a voice of each of multiple speakers within each section from the one or more multi-speaker sections, by using the speaker feature value for each individual speaker and a learned artificial neural network.

Inventors:
YOO TAE JOON (KR)
JOE MYUN CHUL (KR)
CHOI HONG SEOP (KR)
Application Number:
KR2020/008470
Publication Date:
January 07, 2021
Filing Date:
June 29, 2020
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
MINDS LAB INC (KR)
International Classes:
G10L17/04; G06N3/08; G10L17/02; G10L17/18; G10L21/0272
Attorney, Agent or Firm:
Y.P.LEE, MOCK & PARTNERS (KR)
Download PDF: