ISSN: 2229-371X

All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

CONSTRUCTING OF PSEUDOINVARIANTS AND DIGITAL WATERMARKING OF SPEECH SIGNALS BASED ON A SINGULAR VALUE DECOMPOSITION

Dmytro Dmytrovych Peleshko*1, Yuriy Myronovych Pelekh2, Ivan Viktorovych Izonin3
  1. Publishing Information Technologies, Lviv Polytechnic National University, Lviv, Ukraine
  2. Publishing Information Technologies, Lviv Polytechnic National University, Lviv, Ukraine
  3. Publishing Information Technologies, Lviv Polytechnic National University, Lviv, Ukraine
Corresponding Author: Dmytro Dmytrovych Peleshko, E-mail: dpeleshko@gmail.com
Related article at Pubmed, Scholar Google

Visit for more related articles at Journal of Global Research in Computer Sciences

Abstract

Presented digital watermarking method of speech signals based on a constructing of quasistationary intervals, pseudoinvariants and singular value decomposition. Proposed a solution to the problem of building quasistationary intervals of speech signal. This method is based on the characteristics of singular value decomposition of square matrices of operators defined on the elementary intervals. Feature of these methods is their independence from the model voice tract. Watermarking method not dependent on input signal and watermark itself.

Keywords

digital watermarking, singular value decomposition, speech signal, quasistationary intervals, invariant, topology, the operator.

INTRODUCTION

Task of the presplitting of the speech signal into quasistationary intervals is one of the key problems in complex with speech analysis, transformation and synthesis. The successful solution of this task makes the major impact on the efficiency of solving the problem in general. The speech signal is non-stationary and invariant by its nature, although due to the inertia of articulators (lips, tongue, etc.) it is possible to allocate intervals, spectral compositions of which are very similar. Such intervals called quasistationary. In each case, the results of dividing the signal into quasistationary intervals (segmentation) are subjective, because depends on the selected method of further processing. The variety of audio elements of the speech signal and the tasks of its processing leads to the fact that there is not the optimal, and even, the conventional method of segmentation.
Since the speech signal has the property of invariance means that it does not depend on the coordinate system and is often distorted or noisy. The problem of data mining can be reduced to the construction of the fundamental spaces of invariants through which can be divided two non-equal objects.
Analysis of the literature [1] shows that for each of the selected models of voice tract and for each signal processing task can be recommended various methods of initial splitting. Typically, these techniques are closely related to the model voice tract and, accordingly, are not universal.

TASK

The main goal is to develop methods of digital watermarking and segmentation based on characteristics of singular value decomposition using the operator specified in the topological space of speech signal. A feature of these methods is their independence from the signal model. Under the segmentation means the construction of quasistationary intervals.
image
image
image
image
image
There is a white spot (speech signal item) in the center circled by green and red circles what means that specific bit was marked by the watermark and then the same (correct) bit was unmarked.

CONCLUSION

To the decision of segmentation method selection also affects subsequent task of speech signal processing. It is possible that further processing algorithms do not need such detailed segmentation.
Method described in this paper was tested on different words and phrases of speech signal. The experimental results confirm the possibility of using the proposed method for the initial segmentation of speech signal. This method can effectively extend other methods of segmentation based on selected method of voice tract. Building of quasistationary intervals are more efficient then just direct splitting or segmentation of speech signal. In addition, this method improved quality of digital watermarking and made it more robust.

ACKNOWLEDGMENT

We would like to thank our colleagues Yuriy Rashkevych and Anatoliy Kovalchuk for their inputs.

References

  1. Wall, Michael E., Andreas Rechtsteiner, Luis M. Rocha, "Singular value decomposition and principal component analysis". In D.P. Berrar, W. Dubitzky, M. Granzow. A Practical Approach to Microarray Data Analysis. Norwell, MA: Kluwer. – 2003. – С. 91–109.
  2. D. Peleshko, A. Kovalchuk, Y. Pelekh, M. Peleshko Singular decomposition in a speech signal processing. Computer Science And Information Technologies: Materials of the Vith International Scientific and Technical Conference CSIT 2001. - Lviv: Publishig House Vezha&Co, 2011. 2011, c.19-20.
  3. Y. Rashkevych, D. Peleshko, A. Kovalchuk, M. Kupchak, Y. Pelekh Speech signal pseudo invariants. Computer Science And Information Technologies: Materials of the Vith International Scientific and Technical Conference CSIT 2001. - Lviv: Publishig House Vezha&Co, 2011. 2011, c.21-22.
  4. Roger Penrose, “A generalized inverse for matrices.” Proceedings of the Cambridge Philosophical Society 51, 406-413 (1955).
  5. DmytroPeleshko, YuriyPelekh, Ivan Izonin, "Digital Watermarking of Speech Signals", CSIT’2012, p.108, November 2012.
  6. DmytroPeleshko, YuriyPelekh, Ivan Izonin, "Digital Watermarking of Speech Signals Based On Quasistationary Areas", CADS’2013, p.283, 19-23 February 2013.