METHOD FOR FINE-TUNING SINGLE SPEAKER AND THE SYSTEM THEREOF

In a method for fine-tuning a single speaker with a global style token (GST) learning method, the present invention comprises: a step of generating a multi-speaker pretrained model; a step of extracting a single speaker, which is the data of a speaker, from speech data; and a step of individually fi...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	SEUNGKWON LEE, YONGWOO KIM, YOONCHEOL JU
Format:	Patent
Sprache:	eng ; kor
Schlagworte:	ACOUSTICS CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING COUNTING DEAF-AID SETS ELECTRIC COMMUNICATION TECHNIQUE ELECTRICITY LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKEACOUSTIC ELECTROMECHANICAL TRANSDUCERS MUSICAL INSTRUMENTS PHYSICS PUBLIC ADDRESS SYSTEMS SPEECH ANALYSIS OR SYNTHESIS SPEECH OR AUDIO CODING OR DECODING SPEECH OR VOICE PROCESSING SPEECH RECOGNITION
Online-Zugang:	Volltext bestellen
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Beschreibung
Zusammenfassung:	In a method for fine-tuning a single speaker with a global style token (GST) learning method, the present invention comprises: a step of generating a multi-speaker pretrained model; a step of extracting a single speaker, which is the data of a speaker, from speech data; and a step of individually fine-tuning the single speaker. Therefore, the present invention is capable of improving a quality of a model. 본 발명은 GST(Global Style Token) 학습 방식으로 싱글 스피커(single speaker)를 파인튜닝(fine-tuning)하는 방법에 있어서, 멀티 스피커 사전학습된 모델(multi-speaker pretrained model)을 생성하는 단계, 음성 데이터에서 화자의 데이터인 싱글 스피커(single speaker)를 추출하는 단계 및 상기 싱글 스피커를 개별적으로 파인튜닝(fine-tuning)하는 단계를 포함한다.