Performance Evaluation of Novel Digital Audio Watermarking Technique designed for Copyright Protection Application

Performance Evaluation of Novel Digital Audio Watermarking Technique designed for Copyright Protection Application

Ms. Anupama Barai 1, Ms. Sukanya Padave 2,

1, 2 Department of Electronics and Telecommunication K.C.C.E.M.S.R., Thane, India

1 anu4687@gmail.com

2 padave.sukanya@gmail.com

Abstract Online distribution of digital media including images, audio, video and documents has proliferated rapidly in recent years due to large volume of the mobile phones. These files include diverse forms of multimedia such as music, video, text and image. However digital files can be easily copied distributed and altered leading to copyright infringement. Many people download and compress music from the internet creating exact copies of original data. It is this ease of reproducing that causes copyright violations. Copyright protection is the mechanism that prevents data usually digital data from being copied by some unauthorized means. Composers and distributors are more focused on implementing digital watermarking techniques to protect their material against illegal copying.

In this paper a robust watermark system, using spread spectrum with perceptual masking designed for copyright protection application against attack is explained and compared with DCT and 1 D DWT.

KeywordsCopyright protection, Digital audio watermarking, Perceptual masking, Psychoacoustic Auditory Model, Spread Spectrum.

1. INTRODUCTION

   Digital watermarking is the process that embeds copyright information as watermark into the multimedia object, so that the watermark can be extracted to make an assertion about the ownership [7]. The general schematic diagram of watermarking is shown in figure 2 a) and b).

   Copyright Protection can implemented using Digital watermarking [3]. The need of Copyright Protection is explained in Figure 1. Digital watermarking is the process that embeds copyright information as watermark into the multimedia object, so that the watermark can be extracted to make an assertion about the ownership. Selection of Technique: as working on Audio Psychoacoustic Analysis is required for transparent embedding process. Spread Spectrum to enhance the robustness.

2. DIGITAL AUDIO WATERMARKING SYSTEM

   There are many systems for watermarking [4], [5] and [6]. The proposed system comprised of two main steps: first, the watermark generation and embedding and second, the

   watermark recovery [1]. Following Figures 3 and 4 shows the proposed watermarking system.

   Figure. 1. Introduction.

   Figure. 2. a) General Schematic Watermarking System part I.

   1. Watermark Generation and Embedding

      The perceptual property of Human Auditory System is that components below threshold will be imperceptible [2], [8], [9]. In order to analyze the audio signal and compute the amount of masking effect, psychoacoustic model [12], [13] is developed accordingly. In audio watermarking, the noise produced should be inaudible to human ears; this is controlled by psychoacoustic model [10], [11]. In spread spectrum watermarking the watermark is added to the host like an additive noise. To prevent distortions being perceivable,

      amplitude shaping by the minimum masking threshold from this model should be adopted.

      Figure. 2. b) General Schematic Watermarking System part II.

      Spread spectrum is a means of transmission in which the signal occupies a bandwidth in excess of the minimum necessary to send the information; the band spread is accomplished by means of a code which is independent of the data, and a synchronized reception with the code at the receiver is used for de-spreading and subsequent data recovery.

      The process of watermark embedding can be viewed as intention jamming of the watermark signal with the audio signal. In this case the signal (watermark) has much less power than the jammer (audio). It is one of the problems to be overcome at the receiver end. The following chapter expresses the process of watermark generation in spread spectrum terminology. The approach selected in this algorithm is the Direct Sequence Spreading. The system modulates input the data bit-stream with the help of Pseudorandom Number sequence and modulator signal which is usually a cosine function of time with some centre frequency. Figure 5 shows the generation of watermark technique [16].

      The embedding procedure involves the following steps:

      1. The final masking threshold information (as shown in figure 6 and 7) will be used to shape the watermark and embed it into the audio (Psychoacoustic Analysis). Shaping factor is very important for proper embedding.

         Watermark shaping and embedding

         Watermark generation DS/BPSK

         40

         20

         0

         -20

         dB

      2. Shaped Watermark is added in the audio file as an additive noise. Final output is the watermarked audio signal.

   2. Watermark Extraction

      10110…100

      Water

      Marked Audio

      De-spreading And Recovery

      While doing extraction again the psychoacoustic model is referred. Original and Watermarked signal are matched and watermark is extracted. Further Extracted Watermark and original is compared and quantified through correlation coefficient. Further performance evaluation is done against attacks namely Low pass filtering, noising, resampling, requantization and cropping.

      Auditory Model

      Residual Generation

      High resolution Detection

      Recovered watermark

      Figure. 4. Watermark Extraction

      Figure. 5. Watermark Generation.

      Masking components and masking thresholds.

      100

      80

      60

      X: 188

      Y: 43.77

      Frequency index

      Watermark

      (Bit stream)

      Water

      Marked Audio

      Audio

      Fig. 3 Watermark generation and embedding

      P A Model

      250

      200

      150

      100

      50

      0

      Figure.6. Masking Components and Masking Thresholds Using Psychoacoustic Analysis.

      Minimum masking threshold

      75

      70

      65

      60

      55

      50

      45

      40

      X: 17

      Y: 43.77

      Subband number

      35

      30

      25

      20

      15

      10

      5

      0

      dB

      Figure. 7. Minimum Masking threshold Using Psychoacoustic Analysis.

      Figure. 9. Watermarked Audio output of different watermarking system.

3. OBSERVATIONS AND PERFORMANCE EVALUATION

   Whenever the signal recorded in silence it will watermarked efficiently i.e. it will be less prone to attacks. Otherwise it is not getting properly analyzed as important part might be getting discarded under psychoacoustic analysis (if grouped under non-tonal component).

   Figure. 8. System GUI

   Also if signal has pauses in between then it is getting watermarked efficiently. The perceptual quality of the audio signal was retained. The watermarked signal survives to different attacks. Figure 8 shows the System GUI [14], [15]. Figure. 9 shows the watermarked audio output of different watermarking system compared. Figure.10 shows the extracted watermarks through different techniques. Figure 11 shows the extracted watermarks after different attacks with different techniques.

   Figure. 10. Extracted Watermarks

   Figure. 11. Extracted Watermarks after different attacks with different techniques.

4. CONCLUSION

1. The Psychoacoustic model for perceptual masking determination is important for transparent and efficient digital audio watermarkembedding.

2. For a good and efficient watermarking system Robustness and Imperceptibility are two major factors.

3. The outcome depends on type of voice recorded under different conditions (silent or noisy surroundings).

4. The duration for recorded voice should be more than 5 seconds.

5. The robustness was checked on the basis of attacks: cropping, filtering, noising, resampling and requantization.

6. The Watermarking system is proved to be quite robust and imperceptible. The recovery of the watermark was around 99% or almost 100 % for various attacks except for LPF. LPF attack showed variations depending on the type of recorded voice. Retrieval of embedded watermark from the file is more efficiently done (almost 100%) through proposed system than DCT and DWT. In comparison to other techniques like DCT and DWT, the proposed technique showed better results especially for low pass filtering attack and AWGN attack.

ACKNOWLEDGMENT

We would like to extend our thanks to all those who supported and helped to make this work a success.

REFERENCES

1. Ms. Anupama Barai, Mrs. Rohini Deshpande, Copyright Protection by Digital Audio Watermarking using Improved Spread Spectrum with perceptual masking, International Journal of Global Technology Initiatives(IJGTI) volume 2 Issue 1 , March 2013.

2. A Mitchell D. Swanson, Bin Zhu, Ahmed H. Tewfik, Laurence Boney, Robust audio watermarking using perceptual masking ,Signal Processing 66 (1998) pp337-355, 1998 Elsevier Science B.V.

3. Pranab Kumar Dhar, Jong-Myon Kim, Digital Watermarking Scheme Based on Fast Fourier Transformation for Audio Copyright Protection, International Journal of Security and Its Applications Vol. 5 No. 2,.pp33- 48, April, 2011

4. Wahid Barkouti, Lotfi Salhi and Adnan Chérif, Digital audio watermarking using Psychoacoustic model and CDMA modulation, Signal & Image Processing : An International Journal (SIPIJ) Vol.2,

   No.2, June 2011

5. Nedeljko Cvejic, Algorithms for audio watermarking and steganography, University of Oulu., 2004, ISBN 951-42-7383-4.

6. Wu M & Liu B, Multimedia Data Hiding. Springer Verlag, 2003, New York, NY.

7. Arnold M, Wolthusen S & Schmucker M, Techniques and Applications of Digital Watermarking and Content Protection 2003 Artech House, Norwood, MA.

8. Hugo Fastl, Psychoacoustic basis of sound quality evaluation and sound engineering, the 13th International Conference on sound and vibration, July 2-6, 2006, Vienna, Austria

9. David Martin Howard, James A. S. Angus, Acoustics and Psychoacoustics, 4th Edition, illustrated, Focal 2009.

10. Hugo Fastl, Eberhard Zwicker, Psychoacoustics: Facts and Models, Edition 3, Springer 2007.

11. Karl T. Kalveram, Chapter 19, General Introduction to Human Hearing and Speech (pp. 271-276), Part III Human Hearing and Speech of Handbook of Noise and Vibration Control: Malcolm J. Crocker (ed),. Wiley, New Jersey, 2007.

12. Daniel Pressnitzer, Auditory signal processing: physiology, psychoacoustics, and models, International Symposium on Hearing (13th; 2003; Dourdan, France)

13. William A. Yost, Chapter 21 Hearing Thresholds, Loudness of Sound, and Sound Adaptation (pages 286-292), Part III Human Hearing and Speech of Handbook of Noise and Vibration Control: Malcolm J. Crocker (ed),. Wiley, New Jersey, 2007.

14. Oppenheim, A.V., and Schafer, R.W., Discrete-Time Signal Processing (Englewood Cliffs, 1989, NJ: Prentice-Hall).

15. Howard A. Gaberson, A Comprehensive Windows Tutorial, Oxnard,

    California

16. Taub and Schilling, ' Principles of Communication System ', Tata McGraw-Hill, New Delhi, 1995.