Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD FOR SPEECH GENERATION AND RELATED DEVICE
Document Type and Number:
WIPO Patent Application WO/2024/012040
Kind Code:
A1
Abstract:
A method for speech generation and a related device, the method includes: obtaining a first source data input to a speech generation model (100) including multiple encoders(111,112) and a decoder(120)(701), where types of input data of the multiple encoders(111,112) are different; generating a first acoustic feature by a first encoder among the multiple encoders(111,112) based on the first source data(702); and converting a second acoustic feature determined from the first acoustic feature into a third acoustic feature by the decoder(120),where the third acoustic feature is configured to generate a speech with a target voice (703). The above technical solution can rely on a single model to perform different speech generation tasks.

More Like This:
Inventors:
TASNIMA SADEKOVA (CN)
POPOV VADIM (CN)
VLADIMIR GOGORYAN (CN)
MIKHAIL KUDINOV (CN)
Application Number:
PCT/CN2023/094275
Publication Date:
January 18, 2024
Filing Date:
May 15, 2023
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G10L13/047; G10L13/02; G10L13/08
Foreign References:
CN114724541A2022-07-08
CN112233645A2021-01-15
CN112786005A2021-05-11
CN113628610A2021-11-09
KR20080034819A2008-04-22
US6560283B12003-05-06
Attorney, Agent or Firm:
LONGSUN LEAD IP LTD. (CN)
Download PDF: