DNN-based speaker-adaptive postfiltering with limited adaptation data for statistical speech synthesis systems

Name: DNN-based speaker-adaptive postfiltering with limited adaptation data for statistical speech synthesis systems
Author: Öztürk, M. G., Ulusoy, O., Demiroğlu, Cenk

İsim	DNN-based speaker-adaptive postfiltering with limited adaptation data for statistical speech synthesis systems
Yazar	Öztürk, M. G., Ulusoy, O., Demiroğlu, Cenk
Basım Tarihi:	2019
Basım Yeri	- IEEE
Konu	Speaker adaptation, Speech synthesis, Postfilter, Deep learning
Tür	Belge
Dil	İngilizce
Dijital	Evet
Yazma	Hayır
Kütüphane:	Özyeğin Üniversitesi
Demirbaş Numarası	978-1-4799-8131-1
Kayıt Numarası	e71c62e2-a7a5-42e2-9f26-22a7d09970d2
Lokasyon	Electrical & Electronics Engineering
Tarih	2019
Notlar	TÜBİTAK
Örnek Metin	Deep neural networks (DNNs) have been successfully deployed for acoustic modelling in statistical parametric speech synthesis (SPSS) systems. Moreover, DNN-based postfilters (PF) have also been shown to outperform conventional postfilters that are widely used in SPSS systems for increasing the quality of synthesized speech. However, existing DNN-based postfilters are trained with speaker-dependent databases. Given that SPSS systems can rapidly adapt to new speakers from generic models, there is a need for DNN-based postfilters that can adapt to new speakers with minimal adaptation data. Here, we compare DNN-, RNN-, and CNN-based postfilters together with adversarial (GAN) training and cluster-based initialization (CI) for rapid adaptation. Results indicate that the feedforward (FF) DNN, together with GAN and CI, significantly outperforms the other recently proposed postfilters.
DOI	10.1109/ICASSP.2019.8683714

Kaynağa git Özyeğin Üniversitesi

Aramaya Dön

Özyeğin Üniversitesi

Kaynağa git

DNN-based speaker-adaptive postfiltering with limited adaptation data for statistical speech synthesis systems

Yazar Öztürk, M. G., Ulusoy, O., Demiroğlu, Cenk

Basım Tarihi 2019

Basım Yeri - IEEE

Konu Speaker adaptation, Speech synthesis, Postfilter, Deep learning

Tür Belge

Dil İngilizce

Dijital Evet

Yazma Hayır

Kütüphane Özyeğin Üniversitesi

Demirbaş Numarası 978-1-4799-8131-1

Kayıt Numarası e71c62e2-a7a5-42e2-9f26-22a7d09970d2

Lokasyon Electrical & Electronics Engineering

Tarih 2019

Notlar TÜBİTAK

Örnek Metin Deep neural networks (DNNs) have been successfully deployed for acoustic modelling in statistical parametric speech synthesis (SPSS) systems. Moreover, DNN-based postfilters (PF) have also been shown to outperform conventional postfilters that are widely used in SPSS systems for increasing the quality of synthesized speech. However, existing DNN-based postfilters are trained with speaker-dependent databases. Given that SPSS systems can rapidly adapt to new speakers from generic models, there is a need for DNN-based postfilters that can adapt to new speakers with minimal adaptation data. Here, we compare DNN-, RNN-, and CNN-based postfilters together with adversarial (GAN) training and cluster-based initialization (CI) for rapid adaptation. Results indicate that the feedforward (FF) DNN, together with GAN and CI, significantly outperforms the other recently proposed postfilters.

DOI 10.1109/ICASSP.2019.8683714