ELECTRONIC DEVICE, INTELLIGENT SERVER, AND SPEAKER ADAPTIVE SPEECH RECOGNITION METHOD

WIPO Patent Application WO/2024/029875

A1

An electronic device according to an embodiment may comprise: a memory that stores at least one instruction; and at least one processor that is operatively connected to the memory and is for executing the at least one instruction. The at least one processor may perform speaker verification of a speech input to determine whether or not the speech input matches speech of a registered speaker. The at least one processor may perform first speech recognition of the speech input on the basis of a first ASR model if the speech input does not match the speech of the registered speaker. The at least one processor may perform second speech recognition of the speech input on the basis of an SSN and a second ASR model if the speech input matches the speech of the registered speaker. Another embodiment may be possible.

LEE TAEWOO (KR)
KWON MINSEOK (KR)
KIM KYUNGTAE (KR)
SONG GAJIN (KR)
SHIN HOSEON (KR)
LEE JUNGIN (KR)
JUNG SEOKYEONG (KR)

PCT/KR2023/011151

February 08, 2024

July 31, 2023

Click for automatic bibliography generation  

SAMSUNG ELECTRONICS CO LTD (KR)

G10L15/22; G06N3/02; G10L15/26; G10L15/30; G10L17/04; G10L17/06

KR101209033B1	2012-12-06
JP2021039216A	2021-03-11
KR20200109841A	2020-09-23
KR20200120765A	2020-10-21

VESELY KAREL; WATANABE SHINJI; ZMOLIKOVA KATERINA; KARAFIAT MARTIN; BURGET LUKAS; CERNOCKY JAN HONZA: "Sequence summarizing neural network for speaker adaptation", 2016 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP), IEEE, 20 March 2016 (2016-03-20), pages 5315 - 5319, XP032901618, DOI: 10.1109/ICASSP.2016.7472692

MUHANN PATENT & LAW FIRM (KR)

View/Download PDF      PDF Help

Previous Patent: BATTERY MODULE AND BATTERY PACK WITH FIRE EXTINGUISHING FUNCTION

Next Patent: MODIFIED FUSION PROTEIN AND USE THEREOF