Title:
VOICE RECOGNITION DEVICE USING ARTIFICIAL INTELLIGENCE BASED ON TWO-STAGE PRE-LEARNING
Document Type and Number:
WIPO Patent Application WO/2024/071729
Kind Code:
A1
Abstract:
A voice recognition device of the present invention may comprise: a first pre-learning unit, which uses a first artificial intelligence model so as to extract a first feature expression from learning voice data; a second pre-learning unit, which uses a second artificial intelligence model so as to extract a second feature expression from the first feature expression; and a voice recognition unit, which uses a third artificial intelligence model so as to extract predictive text from the second feature expression.
More Like This:
Inventors:
JUNG HO-YOUNG (KR)
KIM JUNE-WOO (KR)
KIM JUNE-WOO (KR)
Application Number:
PCT/KR2023/013475
Publication Date:
April 04, 2024
Filing Date:
September 08, 2023
Export Citation:
Assignee:
KYUNGPOOK NAT UNIV IND ACADEMIC COOP FOUND (KR)
International Classes:
G10L15/26; G06N20/00; G10L15/02; G10L15/06; G10L15/16; G10L15/183; G10L25/18
Foreign References:
| KR20170050029A | 2017-05-11 | |||
| KR20210087099A | 2021-07-09 | |||
| KR20200126675A | 2020-11-09 | |||
| KR20220011065A | 2022-01-27 |
Other References:
JUNE-WOO KIM․HO-YOUNG JUNG*: "End-to-end speech recognition models using limited training data", PHONETICS AND SPEECH SCIENCES, KOREA, vol. 12, no. 4, 1 January 2020 (2020-01-01), Korea, pages 63 - 71, XP009553581, ISSN: 2005-8063
Attorney, Agent or Firm:
KANG, Il Shin et al. (KR)
Download PDF: