ELECTRONIC DEVICE FOR GENERATING PERSONALIZED TTS MODEL AND CONTROL METHOD THEREFOR

• Main Authors: LEE, Jihwan, LEE, Jihyun, MUN, Seongkyu, LEE, Jounyeop, BAE, Jaesung
• Format: Patent
• Language: English; French; Korean
• Published: 06.09.2024
• Subjects: ACOUSTICS; MUSICAL INSTRUMENTS; PHYSICS; SPEECH ANALYSIS OR SYNTHESIS; SPEECH OR AUDIO CODING OR DECODING; SPEECH OR VOICE PROCESSING; SPEECH RECOGNITION

This control method for an electronic device comprises the steps of: acquiring second speech data, including information about speech spoken by a user, in order to personalize a text-to-speech (TTS) model trained on the basis of first speech data including information about speech spoken by a plurality of speakers; generating training data by combining one piece of the first speech data and one piece of the second speech data; and using the training data to train the TTS model. Ce procédé de commande pour un dispositif électronique comprend les étapes consistant à : acquérir des secondes données vocales, comprenant des informations concernant la parole prononcée par un utilisateur, afin de personnaliser un modèle texte-parole (TTS) entraîné sur la base de premières données vocales comprenant des informations concernant la parole prononcée par une pluralité de haut-parleurs ; générer des données d'apprentissage en combinant un élément des premières données vocales et un élément des secondes données vocales ; et utiliser les données d'apprentissage pour entraîner le modèle TTS. 전자 장치의 제어 방법은 복수의 화자가 발화한 음성에 대한 정보를 포함하는 제1 음성 데이터에 기초하여 학습된 TTS(Text-To-Speech) 모델을 개인화하기 위하여, 사용자가 발화한 음성에 대한 정보를 포함하는 제2 음성 데이터를 획득하는 단계, 제1 음성 데이터 중 하나와 제2 음성 데이터 중 하나를 결합하여 학습 데이터를 생성하는 단계 및 학습 데이터를 이용하여 TTS 모델을 학습시키는 단계를 포함한다.