High-quality voice conversion system based on 
GMM statistical parameters and RBF neural network

doi:10.1016/S1005-8885(14)60333-2

中国邮电高校学报(英文) ›› 2014, Vol. 21 ›› Issue (5): 68-75.doi: 10.1016/S1005-8885(14)60333-2

High-quality voice conversion system based on GMM statistical parameters and RBF neural network

陈先同

南京邮电大学

收稿日期:2014-01-16 修回日期:2014-06-24 出版日期:2014-10-31 发布日期:2014-10-30
通讯作者: 陈先同 E-mail:chenxt0524@126.com
基金资助:
国家自然科学基金

High-quality voice conversion system based on GMM statistical parameters and RBF neural network

CHEN Xiantong

Received:2014-01-16 Revised:2014-06-24 Online:2014-10-31 Published:2014-10-30
Contact: CHEN Xiantong E-mail:chenxt0524@126.com

摘要/Abstract

摘要： 　A voice conversion (VC) system was designed based on Gaussian mixture model (GMM) and radial basis function (RBF) neural network. As a voice conversion model, RBF network needs quantities of training data to improve its performance. For one speech, the networks trained by different segments of data have different transformation effects. Since trying segment by segment to obtain the best conversion effect is complex, a conversion method was proposed, that uses GMM for statistics before training RBF network to aim at the problem. The speech transformation and representation using adaptive interpolation of weighted spectrum (STRAIGHT) model is used for accurate extraction of vocal tract spectrum. Then GMM is used to classify the numerous spectral parameters. The obtained mean parameters were trained in RBF network. Experiment reveals that, the soft classification ability of GMM can promptly realize the reduction and classification of training data under the premise of ensuring the training effect. The selection complexity is decreased thereafter. Compared to the conventional RBF network training methods, this method can make the transformation of spectral parameters more effective and improve the quality of converted speech.

关键词: VC system, STRAIGHT, vocal tract spectrum, GMM, RBF

Abstract: 　A voice conversion (VC) system was designed based on Gaussian mixture model (GMM) and radial basis function (RBF) neural network. As a voice conversion model, RBF network needs quantities of training data to improve its performance. For one speech, the networks trained by different segments of data have different transformation effects. Since trying segment by segment to obtain the best conversion effect is complex, a conversion method was proposed, that uses GMM for statistics before training RBF network to aim at the problem. The speech transformation and representation using adaptive interpolation of weighted spectrum (STRAIGHT) model is used for accurate extraction of vocal tract spectrum. Then GMM is used to classify the numerous spectral parameters. The obtained mean parameters were trained in RBF network. Experiment reveals that, the soft classification ability of GMM can promptly realize the reduction and classification of training data under the premise of ensuring the training effect. The selection complexity is decreased thereafter. Compared to the conventional RBF network training methods, this method can make the transformation of spectral parameters more effective and improve the quality of converted speech.

Key words: VC system, STRAIGHT, vocal tract spectrum, GMM, RBF

CHEN Xiantong. High-quality voice conversion system based on GMM statistical parameters and RBF neural network[J]. Acta Metallurgica Sinica(English letters), 2014, 21(5): 68-75.

参考文献

1. Jian Z H, Yang Z. An overview of voice conversion. Journal of Nanjing University of Posts and Telecommunications, 2007, 27 (6): 88-94 (in Chinese)

2. Abe M. A segment-based approach to voice conversion. IEEE International Conference on Acoustics, Speech and Signal Processing, May 14-17, 1991, Toronto, Canada. Piscataway, NJ, USA: IEEE, 1991: 765-768

3. Mizuno H, Abe M. Voice conversion algorithm based on piecewise linear conversion rules of formant frequency and spectrum tilt. Speech Communication, 1995, 16(2): 153-164

4. Narendranath M, Murthy H A, Rajendran S, et al. Transformation of formants for voice conversion using artificial neural networks. Speech Communication.,1995, 16(2): 207-216

5. Zuo G Y, Liu W J, Ruan X G. Genetic algorithm based RBF neural network for voice conversion. Proceedings of the World Congress on Intelligent Control and Automation (WCICA’04): Vol 5, Jun 15-19, 2004, Hangzhou, China. Piscataway, NJ, USA: IEEE, 2004: 4215-4218

6. Kawahara H. STRAIGHT, exploitation of the other aspect of VOCODER: Perceptually isomorphic decomposition of speech sounds. Acoustical Science and Technology, 2006, 27 (6): 349-353

7. Girin L. Long-term quantization of speech LSF parameters. Proceedings of the 2007 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP’07): Vol.4, Apr 15-20, 2007, Honolulu, HI, USA. Piscataway, NJ, USA: IEEE, 2007: 845-848

8. Mouchtaris A, Agiomyrgiannakis Y, Stylianou Y. Conditional vector quantization for voice conversion. Proceedings of the 2007 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP’07): Vol.4, Apr 15-20, 2007, Honolulu, HI, USA. Piscataway, NJ, USA: IEEE, 2007: 505-508

9. Xie W C, Zhang L H. Research on voice conversion based codebook and GMM. Proceedings of the IEEE 12th International Conference on Communication Technology (ICCT’10), Nov 11-14, 2010, Nanjing, China. Piscataway, NJ, USA: IEEE, 2010: 1043-1046

10. Makki B, Seyedsalehi S A, Sadati N, et al. Voice conversion using nonlinear principal component analysis. Proceedings of the 2007 IEEE Symposium on Computational Intelligence in Image and Signal Processing (CIISP’07), Apr 1-5, 2007, Honolulu, HI, USA. Piscataway, NJ, USA: IEEE, 2007: 336-339

11. Qiao Y, Minematsu N. Mixture of probabilistic linear regressions: A unified view of GMM-based mapping techniques. Proceedings of the 2009 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP’09), Apr 19-24, 2009, Taipei, China. Piscataway, NJ, USA: IEEE, 2009: 3913-3916

12. Chen Z, Zhang L H. Morphing arithmetic of pitch and application in voice morphing system. Journal of Nanjing University of Posts and Telecommunications, 2010, 30 (5): 83-87 (in Chinese)

13. Zhang Z J, Yang W Y, Chen Z. Voice conversion based on STRAIGHT model and artificial neural networks. Audio Engineering, 2010, 34 (9): 49-52 (in Chinese)

14. Godoy E, Rosec O, Chonacel T. Voice conversion using dynamic frequency warping with amplitude scaling, for parallel or nonparallel corpora. IEEE Transactions on Audio, Speech and Language Processing, 2012, 20 (4): 1313-1323

15. Daniel E, Asuncion M, Antonio B. Voice conversion based on weighted frequency warping. IEEE Transactions on Audio, Speech and Language Processing, 2010, 18 (5): 922-931

High-quality voice conversion system based on GMM statistical parameters and RBF neural network

High-quality voice conversion system based on GMM statistical parameters and RBF neural network

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics

本文评价