APPARATUS METHOD AND COMPUTER PROGRAM FOR RECOGNIZING VOICE OF CHILD

An apparatus for recognizing the speech of a child comprises: a feature vector extracting unit for extracting a feature vector of adult speech from adult speech data; a warping factor value application unit for estimating a warping factor value from the extracted feature vector of the adult speech b...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: CHOI BO KYEONG
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:An apparatus for recognizing the speech of a child comprises: a feature vector extracting unit for extracting a feature vector of adult speech from adult speech data; a warping factor value application unit for estimating a warping factor value from the extracted feature vector of the adult speech based on a pre-learned single phoneme acoustic model and applying the estimated warping factor value to the adult speech data; a learning unit for learning a speaker adaptive acoustic model based on the adult speech data to which the warping factor value is applied; a speech input unit which receives a voice spoken by a child; and a speech recognition unit for recognizing the speech of the child from the input child speech data based on the speaker adaptive acoustic model. According to the present invention, the age range of speech recognition is extended from adults to children. 어린이 음성을 인식하는 장치는 성인 음성 데이터로부터 성인 음성의 특징 벡터를 추출하는 특징 벡터 추출부, 기학습된 단일 음소 음향 모델에 기초하여 상기 추출된 성인 음성의 특징 벡터로부터 워핑 요소(warping factor) 값을 추정하고, 상기 추정된 워핑 요소 값을 상기 성인 음성 데이터에 적용하는 워핑 요소 값 적용부, 상기 워핑 요소 값이 적용된 상기 성인 음성 데이터에 기초하여 화자 적응 음향 모델을 학습시키는 학습부, 어린이로부터 발화된 음성을 입력받는 음성 입력부 및 상기 화자 적응 음향 모델에 기초하여 상기 입력된 어린이 음성 데이터로부터 상기 어린이 음성을 인식하는 음성 인식부를 포함한다.