METHOD COMPUTER DEVICE AND COMPUTER PROGRAM FOR SPEAKER DIARIZATION COMBINED WITH SPEAKER IDENTIFICATION

Disclosed are a method, system, and computer program for separating a speaker combined with a speaker identification. The method for separating the speaker comprises: a step of setting a reference speech in relation to a speech file received as a target speech for speaker separation from a client; a...

Full description

Saved in:
Bibliographic Details
Main Authors HAN ICKSANG, KANG HAN YONG, HEO HEE SOO, KIM YOU JIN, KWON YOUNGKI, LEE BONG JIN, JANG JUNGHOON, KIM HAN GYU, CHUNG JOON SON
Format Patent
LanguageEnglish
Korean
Published 22.07.2022
Subjects
ACOUSTICS
MUSICAL INSTRUMENTS
PHYSICS
SPEECH ANALYSIS OR SYNTHESIS
SPEECH OR AUDIO CODING OR DECODING
SPEECH OR VOICE PROCESSING
SPEECH RECOGNITION
Online AccessGet full text

Cover

More Information
Summary:Disclosed are a method, system, and computer program for separating a speaker combined with a speaker identification. The method for separating the speaker comprises: a step of setting a reference speech in relation to a speech file received as a target speech for speaker separation from a client; a step of performing speaker identification to identify a speaker of the reference speech in the speech file using the reference speech; and a step of performing speaker separation using clustering for the remaining unidentified utterance sections in the speech file. Therefore, the present invention is capable of improving a speaker separation performance. 화자 식별과 결합된 화자 분리 방법, 시스템, 및 컴퓨터 프로그램이 개시된다. 화자 분리 방법은, 클라이언트로부터 화자 분리 대상 음성으로 수신된 음성 파일과 관련하여 기준 음성을 설정하는 단계; 상기 기준 음성을 이용하여 상기 음성 파일에서 상기 기준 음성의 화자를 식별하는 화자 식별을 수행하는 단계; 및 상기 음성 파일에서 식별되지 않은 나머지 발화 구간에 대해 클러스터링을 이용한 화자 분리를 수행하는 단계를 포함한다.
Bibliography:Application Number: KR20210006190