Automatic Control and Computer Sciences

, Volume 52, Issue 6, pp 561–571 | Cite as

Separation of Reverberant Speech Based on Computational Auditory Scene Analysis

  • Li HongyanEmail author
  • Cao Meng
  • Wang Yue


This paper proposes a computational auditory scene analysis approach to separation of room reverberant speech, which performs multi-pitch tracking and supervised classification. The algorithm trains speech and non-speech model separately, which learns to map from harmonic features to grouping cue encoding the posterior probability of time-frequency unit being dominated by the target and periodic interference. Then, a likelihood ratio test selects the correct model for labeling time-frequency unit. Experimental results show that the proposed approach produces strong pitch tracking results and leads to significant improvements of predicted speech intelligibility and quality. Compared with the classical Jin-Wang algorithm, the average SNR of this algorithm is improved by 1.22 dB.


computational auditory scene analysis room reverberant supervised classification harmonic features 



This work was supported by Shanxi Natural Science Foundation (no. 201701D121058).


  1. 1.
    Mingyang Wu and DeLiang Wang, A two-stage algorithm for one-microphone reverberant speech enhancement, IEEE Trans. Audio Speech Lang. Process., 2006, vol. 14, no. 3, pp. 774–784.CrossRefGoogle Scholar
  2. 2.
    Zhaozhang Jin and DeLiang Wang, A supervised learning approach to monaural segregation of reverberant speech, IEEE Trans. Audio Speech Lang. Process., 2009, vol. 17, no. 4, pp. 625–638.CrossRefGoogle Scholar
  3. 3.
    Cooke, M.P., Modeling Auditory Processing and Organization, Cambridge, UK: Cambridge University Press, 1993.Google Scholar
  4. 4.
    Wei Guo and Fengjin Yu, Speech-music signal separation based on improved time-frequency ratio, Comput. Eng., 2015, vol. 41, no. 3, pp. 287–291.Google Scholar
  5. 5.
    Moore, B.C.J., An Introduction to the Psychology of Hearing, London: Academic Press, 5th ed.Google Scholar
  6. 6.
    Xiaojia Zhao and Yang Shao, CASA-based robust speaker identification, IEEE Trans. Audio Speech Lang. Process., 2012, vol. 20, no. 5, pp. 1608–1616.CrossRefGoogle Scholar
  7. 7.
    Jianfen Ma, Research on Blind Separation and Enhancement of Speech Signals, Beijing: Electronic Industry Press, 2012.Google Scholar
  8. 8.
    Yu Wang, Jiajun Lin, and Wenhao Yuan, Improved speech enhancement based on computational auditory scene analysis, J. East China Univ. Sci. Technol. (Natl. Sci. Ed.), 2012, vol. 38, no. 5, pp. 617–621.Google Scholar
  9. 9.
    Chun Wu, Cochannel Speech Separation Based on Computational Auditory Scene Analysis, Guangxi University, 2014.Google Scholar
  10. 10.
    Qi Hu, Single-Channel Speech Separation Based on Computational Auditory Scene Analysis, Beijing Jiaotong University, 2014.Google Scholar
  11. 11.
    Ubul Kurban, Hamdulla Askar, and Aysa Alim, A digital signal processing teaching methodology using Praat, 2009 4th International Conference on Computer Science and Education, Nanning: IEEE, 2009.Google Scholar
  12. 12.
    Li Hong-yan, Qu Jun-ling, and Zhang Xue-ying, The voiced speech blind signal separation algorithm based on signal energy, J. Jilin Univ. Eng. Technol. Ed., 2015, vol. 45, no. 5, pp. 1665–1670.Google Scholar
  13. 13.
    Liheng Zhao and Zhengfu Wang, Monaural voiced speech separation based on harmonic and energy features, Acta Acust., 2012, vol. 37, no. 2, pp. 218–224.Google Scholar
  14. 14.
    Lehmanna, E.A. and Johansson, A.M., Prediction of energy decay in room impulse responses simulated with an image-source model, Acoust. Soc. Am., 2008, vol. 124, no. 1, pp. 269–277.CrossRefGoogle Scholar
  15. 15.
    Xueliang Zhang, Yiju Liu, and Peng Li, Monaural voiced speech segregation based on improved harmonic grouping rules, Acta Acust., 2011, vol. 36, no. 1, pp. 88–96.Google Scholar

Copyright information

© Allerton Press, Inc. 2018

Authors and Affiliations

  1. 1.College of Information Engineering, Taiyuan University of Technology TaiyuanTaiyuanChina

Personalised recommendations