Blind speech signal separation based on cumulant approach

مهندسی مخابرات جنوب

Article 1, Volume 6, Issue 23, July 2017   

Document Type: Research Article

Authors

Sahar Pouya^{* 1}; m Esmail Beyg¹; R. Hamzehyan²

¹Department of Electrical Engineering, Islamic Azad University, Bushehr Branch, Bushehr, Iran

²Islamic Azad University, Bushehr Branch, Faculty Member, Department of Electrical Engineering

Abstract

One of the proposed methods for separating several speech signals, which are combined in receivers, is the use of blind source separation (BSS) methods. Resource blind separation is the separation and estimation of signals generated by sources in an unknown channel and their combinations received at receivers. Existing algorithms for blind source separation are often based on the special analysis of fourth-order cumulative matrices. However, when cumulative matrices have close eigenvalues, their eigenvectors become very sensitive to the error in estimating the matrices. In this paper, we try to reduce this sensitivity by using a new algorithm and obtain a more accurate estimate.

Keywords

Blind signals; Cumulative; Eigenvector; eigenvalues

References

[1]   P. Comon and L. Rota, “Blind separation of independent sources from convolutive mixtures,” IEICE Trans. on Fundamentals, vol. E86-A, no. 3, pp. 542–549, Mar 2003.

[2]   M. Syskind Pedersen, J. Larsen , U. Kjems and L. C. Parra, (2009), "A survey of convolutive blind source separation methods", Springer Handbook on Speech Processing and Speech Communication, pp. 1-34, 2007.

[3]   R. Martin-clemente and J. I. Acha, "New equations and iterative algorithm for blind separation of sources", Signal Processing, vol. 82, pp. 861-873, 2002.

[4]   A. Hyvarinen, J. Karhunen and E. Oja, (2011), "Independent component analysis", John Wiley & Sons, Inc

[5]   R. Martin-clemente, J. I. Acha and C. G.Puntonet, "Eigendecomposition of self-tuned cumulant-matrices for blind source separation", Signal Processing, vol. 84, pp. 1201- 1211, 2004.