| Improving Deep Neural Network Based Speech Synthesis through Contextual Feature Parametrization and Multi-Task Learning | |
| Wen, Zhengqi1; Li, Kehuang2; Huang, Zhen2; Lee, Chin-Hui2; Tao, Jianhua1,3,4; Zhengqi Wen | |
| Source Publication | JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY
 |
| 2018-07-01 | |
| Volume | 90Issue:7Pages:1025-1037 |
| Subtype | Article |
| Abstract | We propose three techniques to improve speech synthesis based on deep neural network (DNN). First, at the DNN input we use real-valued contextual feature vector to represent phoneme identity, part of speech and pause information instead of the conventional binary vector. Second, at the DNN output layer, parameters for pitch-scaled spectrum and aperiodicity measures are estimated for constructing the excitation signal used in our baseline synthesis vocoder. Third, the bidirectional recurrent neural network architecture with long short term memory (BLSTM) units is adopted and trained with multi-task learning for DNN-based speech synthesis. Experimental results demonstrate that the quality of synthesized speech has been improved by adopting the new input vector and output parameters. The proposed BLSTM architecture for DNN is also beneficial to learning the mapping function from the input contextual feature to the speech parameters and to improve speech quality. |
| Keyword | Dnn-based Speech Synthesis Vocoder Speech Parametrization Blstm Phoneme Embedded Vector Multi-task Learning Pitch-scaled Spectrum |
| WOS Headings | Science & Technology ; Technology |
| DOI | 10.1007/s11265-017-1293-z |
| WOS Keyword | RECOGNITION ; REPRESENTATIONS ; DIVERGENCE ; EXTRACTION ; GENERATION ; SELECTION |
| Indexed By | SCI |
| Language | 英语 |
| Funding Organization | National High-Tech Research and Development Program of China (863 Program)(2015AA016305) ; National Natural Science Foundation of China (NSFC)(61403386) ; Strategic Priority Research Program of the CAS(XDB02080006) ; Major Program for the National Social Science Fund of China(13ZD189) |
| WOS Research Area | Computer Science ; Engineering |
| WOS Subject | Computer Science, Information Systems ; Engineering, Electrical & Electronic |
| WOS ID | WOS:000433555600007 |
| Citation statistics | |
| Document Type | 期刊论文 |
| Identifier | http://ir.ia.ac.cn/handle/173211/19932 |
| Collection | 模式识别国家重点实验室_语音交互 |
| Corresponding Author | Zhengqi Wen |
| Affiliation | 1.Natl Lab Pattern Recognit, Beijing, Peoples R China 2.Georgia Inst Technol, Sch Elect & Comp Engn, Atlanta, GA 30332 USA 3.Chinese Acad Sci, Inst Automat, CAS Ctr Excellence Brain Sci & Intelligence Techn, Beijing, Peoples R China 4.Univ Chinese Acad Sci, Sch Comp & Control Engn, Beijing, Peoples R China |
| Recommended Citation GB/T 7714 | Wen, Zhengqi,Li, Kehuang,Huang, Zhen,et al. Improving Deep Neural Network Based Speech Synthesis through Contextual Feature Parametrization and Multi-Task Learning[J]. JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY,2018,90(7):1025-1037. |
| APA | Wen, Zhengqi,Li, Kehuang,Huang, Zhen,Lee, Chin-Hui,Tao, Jianhua,&Zhengqi Wen.(2018).Improving Deep Neural Network Based Speech Synthesis through Contextual Feature Parametrization and Multi-Task Learning.JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY,90(7),1025-1037. |
| MLA | Wen, Zhengqi,et al."Improving Deep Neural Network Based Speech Synthesis through Contextual Feature Parametrization and Multi-Task Learning".JOURNAL OF SIGNAL PROCESSING SYSTEMS FOR SIGNAL IMAGE AND VIDEO TECHNOLOGY 90.7(2018):1025-1037. |
| Files in This Item: | Download All | |||||
| File Name/Size | DocType | Version | Access | License | ||
| 2017_Journal of Sina（995KB） | 期刊论文 | 作者接受稿 | 开放获取 | CC BY-NC-SA | View Download | |
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment