Blind source separation algorithm for convolution mixed signals

Blind source separation algorithm for convolution mixed signals

Chunli Wang Quanyu Wang  Yuping Cao 

College of Electronics and Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China

Corresponding Author Email:
31 December 2017
| Citation



In the actual speech enhancement application, a large number of observation data need longer filters. The time domain algorithm has the disadvantages of large computation amount and slow processing speed. Transforming the time domain convolution operation into the frequency domain product operation can not only avoid the complicated convolution operation, but also reduce the calculation amount to a large extent, and improve the effectiveness of the blind source separation algorithm. Simulation experiment results show that the blind deconvolution algorithm in the frequency domain can improve the intelligibility and articulation of separated speech.


Speech Enhancement, Frequency Domain, Convolution, Blind Source Separation, Effectiveness

1. Introduction
2. Mathematical Model of Convolution Mixing
3. Frequency Domain Blind Deconvolution Algorithm
4. Simulation Experiment and Analysis of Results
5. Conclusions

[1] Mitianoudis N., Stathaki T. (2007). Batch and online underdetermined source separation using alpaca mixture models, IEEE Transactions on Audio, Speech and Language Processing, Vol. 15, No. 6, pp. 1818-1832.

[2] Pedersen M.S., Wang D.L., Larsen J., Kjems U. (2008). Two- microphones separation of speech mixtures, IEEE Transactions on Neural Networks, Vol. 19, No. 3, pp. 475-492. DOI: 10.1109/TNN.2007.911740

[3] Hiroshi S., Shoko A., Shoji M. (2011). Underdetermined convolutive blind source separation via frequency bin- wise clustering and permutation alignment, IEEE Transactions on Audio, Speech, and Language Processing, Vol. 19, No. 3, pp. 516-527. DOI: 10.1109/TASL.2010.2051355 

[4] Rivet B., Girin L., Jutten C. (2007). Mixing audiovisual speech processing and blind source separation for the extraction of speech signals from convolutive mixtures, IEEE Trans on Audio, Speech and Language Processing, Vol. 15, No. 1, pp. 96-108. DOI: 10.1109/TASL.2006.872619

[5] Kirei B.S., Topa M., Muresan I., Homana I., Toma N. (2011). Blind source separation for convolutive mixtures with neural networks, Advances in Electrical and Computer Engineering, Vol. 11, pp. 63-68. DOI: 10.4316/AECE.2011.01010

[6] Prasad R., Saruwatari H., Shikano K. (2009). Enhancement of speech signals separated from their convolutive mixture by FDICA algorithm, Digital Signal Processing, Vol. 19, pp. 127-133. DOI: 10.1016/j.dsp.2008.01.007

[7] Wang L., Ding H., Yin F. (2010). An improved method for permutation correction in convolutive blind source separation, Archives of Acoustics, Vol. 35, No. 4, pp. 493-504. DOI: 10.2478/v10168-010-0038-9

[8] Guo W., Yu F.Q. (2015). Improved speech music signal separation based on negative entropy maximization, Computer Engineering and Application, Vol. 51, No. 4, pp. 209-212. DOI: 10.3778/j.issn.1002-8331.1306-0039

[9] Zhang Y.Y., Xin J.H., Liu G.B. (2016). Applications of combined with cumulant slice joint diagonalization of blind source separation, Journal of Huazhong University of Science and Technology (Natural Science), Vol. 44, No. 7, pp. 86-90. DOI: 10.13245/j.hust.160717

[10] Zhou J. (2016). Research of underdetermined source estimation and blind extraction method for mechanical fault signals, Doctoral Dissertation of Kunming University, pp. 38-45.

[11] Yang J.M., Qi H.Y. (2015). Improved nonlinear blind source separation algorithm based on the minimization of mutual information, Electric Measurement and Instrument, Vol. 52, No. 9, pp. 66-69. DOI: 10.3969/j.issn.1001-1390.2015.09.013