Advertisement

Mongolian Text-to-Speech System Based on Deep Neural Network

  • Rui Liu
  • Feilong Bao
  • Guanglai Gao
  • Yonghe Wang
Conference paper
Part of the Communications in Computer and Information Science book series (CCIS, volume 807)

Abstract

Recently, Deep Neural Network (DNN), which is a feed-forward artificial neural network with many hidden layers, has opened a new research direction for Speech Synthesis. It can represent high dimension and correlated features efficiently and model highly complex mapping function compactly. However, the research on DNN-based Mongolian speech synthesis is still in blank filed. This paper applied the DNN-based acoustic model to Mongolian speech synthesis firstly, and built a Mongolian speech synthesis system according to the Mongolian character and acoustic features. Compared with the conventional HMM-based system under the same corpus, the DNN-based system can synthesize better Mongolian speech than HMM-based system can do. The Mean Opinion Score (MOS) of the synthesized Mongolian speech is 3.83. And it becomes a new state-of-the-art system in this field.

Keywords

Mongolian Text-to-Speech (TTS) Acoustic model Deep Neural Network (DNN) 

Notes

Acknowledgments

This research was supports in part by the China national natural science foundation (No. 61563040, No. 61773224) and Inner Mongolian nature science foundation (No. 2016ZD06).

References

  1. 1.
    Ethnologue: Languages of the world, 18th edition. http://www.ethnologue.com
  2. 2.
    Bao, F., Gao, G., Yan, X., et al.: Segmentation-based Mongolian LVCSR approach. In: 38th IEEE International Conference on Acoustics, Speech and Signal Processing, Canada, pp. 8136–8139. IEEE Press (2013)Google Scholar
  3. 3.
    Ochir, Zheng, G.: A test of the speech synthesis with the waveform concatenation. In: 3rd National Conference on Man-Machine Speech Communication, Chongqing, pp. 408–412 (1994)Google Scholar
  4. 4.
    Monghjaya: A research on Mongolian speech synthesis system based on stems and affixes. J. Inner Mong. Univ. 39, 693–697 (2008)Google Scholar
  5. 5.
    Aomin, Ziyu, X., He, H., et al.: A study on the piano and rhyme phrases of Mongolian. In: 10th Phonetic Conference of China Processing, Shanghai (2012)Google Scholar
  6. 6.
    Zhao, J., Gao, G., Bao, F.: Research on HMM-based Mongolian speech synthesis. Comput. Sci. 41, 80–104 (2014)Google Scholar
  7. 7.
    Bengio, Y.: Learning deep architectures for AI. Found. Trends Mach. Learn. 2, 1–55 (2009)CrossRefzbMATHGoogle Scholar
  8. 8.
    Hinton, G., Li, D., Yu, D., et al.: Deep neural networks for acoustic modeling in speech recognition: the shared views of four research groups. IEEE Signal Process. Mag. 29, 82–97 (2012)CrossRefGoogle Scholar
  9. 9.
    Zen, H., Senior, A., Schuster, M.: Statistical parametric speech synthesis using deep neural networks. In: 38th IEEE International Conference on Acoustics, Speech and Signal Processing, Canada, pp. 7962–7966. IEEE Press (2013)Google Scholar
  10. 10.
  11. 11.
  12. 12.
    Yoshimura, T., Tokuda, K., Masuko, T., et al.: State duration modeling for HMM-based speech synthesis. IEEE Trans. Inf. Syst. 90, 692–693 (2007)Google Scholar
  13. 13.
    Yoshimura, T., Tokuda, K., Masuko, T., et al.: Simultaneous modeling of spectrum, pitch and duration in HMM-based speech synthesis. In: 6th European Conference on Speech Communication and Technology, Hungary, pp. 2099–2107. IEEE Press (1999)Google Scholar
  14. 14.
    Yan, X., Bao, F., Wei, H., Su, X.: A novel approach to improve the Mongolian language model using intermediate characters. In: Sun, M., Huang, X., Lin, H., Liu, Z., Liu, Y. (eds.) CCL/NLP-NABD -2016. LNCS (LNAI), vol. 10035, pp. 103–113. Springer, Cham (2016).  https://doi.org/10.1007/978-3-319-47674-2_9 CrossRefGoogle Scholar
  15. 15.
    Bao, F., Gao, G., Yan, X.: Research on grapheme to phoneme conversion for Mongolian. Appl. Res. Comput. 30, 1696–1700 (2013)Google Scholar
  16. 16.
    Liu, R., Bao, F., Gao, G., Zhang, H.: Approach to predicition Mongolian prosody phrase based on CRF model. In: 13th National Conference on Man-Machine Speech Communication, Tianjin (2015)Google Scholar
  17. 17.
    Liu, R., Bao, F., Gao, G.: Mongolian prosodic phrase prediction using suffix segmentation. In: 20th International Conference on Asian Language Processing, Taiwan, pp. 250–253. IEEE Press (2016)Google Scholar
  18. 18.
    Tokuda, K., Yoshimura, T., Masuko, T., et al.: Speech parameter generation algorithms for HMM-based speech synthesis. In: 25th IEEE International Conference on Acoustics, Speech, and Signal Processing, Istanbul, pp. 1315–1318. IEEE Press (2000)Google Scholar
  19. 19.
    Milner, B., Shao, X.: Speech reconstruction from mel-frequency cepstral coefficients using a source-filter model. In: 7th International Conference on Spoken Language Processing, Denver. IEEE Press (2002)Google Scholar
  20. 20.
  21. 21.
  22. 22.
    Masuko, T.: Multi-space probability distribution HMM. IEEE Trans. Inf. Syst. 85, 455–464 (2002)Google Scholar
  23. 23.
    Grünwald, P.: The Minimum Description Length Principle, vol. 1, pp. 257–268. MIT Press (2007)Google Scholar
  24. 24.
    Lu, H., Ling, Z.-H., Lei, M., et al.: Minimum generation error based optimization of HMM model clustering for speech synthesis. Pattern Recognit. Artif. Intell. 23, 822–828 (2010)Google Scholar
  25. 25.
    Zen, H., Sak, H.: Unidirectional long short-term memory recurrent neural network with recurrent output layer for low-latency speech synthesis. In: 40th IEEE International Conference on Acoustics, Speech and Signal Processing, Australia, pp. 4470–4474. IEEE Press (2015)Google Scholar
  26. 26.
    Achanta, S., Godambe, T., Gangashetty, S.V.: An investigation of recurrent neural network architectures for statistical parametric speech synthesis. In: 16th Interspeech, Germany. IEEE Press (2015)Google Scholar
  27. 27.
    Wu, Z., King, S.: Investigating gated recurrent networks for speech synthesis. In: 41st IEEE International Conference on Acoustics, Speech and Signal Processing, Shanghai, pp. 5140–5144. IEEE Press (2016)Google Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2018

Authors and Affiliations

  • Rui Liu
    • 1
  • Feilong Bao
    • 1
  • Guanglai Gao
    • 1
  • Yonghe Wang
    • 1
  1. 1.College of Computer ScienceInner Mongolia UniversityHohhotChina

Personalised recommendations