2012 International Conference on Computing, Electronics and Electrical Technologies [ICCEET]

Audio Authentication Using Arnold and Discrete

Cosine Transform
N.V.Lalitha

Gulivindala Suresh

Department of ECE
GMR Institute of Technology
RAJAM-532 127, A.P. INDIA.
E-mail: lalitha.nv@gmrit.org

Prabhakar Telagarapu
Department of ECE
GMR Institute of Technology
RAJAM-532 127, A.P. INDIA.
Email: prabhakar.t@gmrit.org

Department of ECE
GMR Institute of Technology
RAJAM-532 127, A.P. INDIA.
Email: suresh.g@gmrit.org

For u=0,1,2,.,N-1. Similarly, the inverse transform is

defined as,

AbstractDigital Audio Watermarking involves the concealment

of data within a discrete audio file. Digital watermarking is
identified as a partial solution to related problems like illegal
reproduction and distribution of digital media. Intellectual
property protection is the main driving force behind research in
this area. In order to provide authentication the watermark
image is scrambled with Arnold transformation and then
embedded into the Discrete Cosine Transformed (DCT) original
audio signal. The performance of this technique is evaluated in
terms of PSNR, MOS Criteria and Pearson Correlation
Coefficient (PCC).

(2)
For x=0, 1, 2,.N-1. In both equations (1) and (2) (u) is
defined as,

Keywords-Audio watermarking; Arnold Transform; Discrete

Cosine Transform; MOS Criteria; PSNR; Pearson Correlation
Coefficient;

It is clear from eq.(1) that for u=0, C(0)=

I.

0
0

1 1
1 2
III.

(4)
WATERMARK EMBEDDING ALGORITHM

The process of watermark embedding is shown in Fig.1.

Step 1: Sample the original audio signal at a sampling rate (fs)
and partition the sampled file into frames each having certain
samples.
Step 2: Perform DCT transformation on original audio signal.
This operation produces Two sub-bands: A, D. The D
represents the Details sub-band, and A represents the
approximation sub-band.
Step 3: Apply Arnold Transform and DCT to watermark
image.
Step 4: Embed the transformed watermark coefficients into
the DCT transformed original audio signal.
Step 5: Apply the inverse DCT operation to obtain each
watermarked audio frame.

A. The DCT transform

The Discrete Cosine Transform is a technique for
converting a signal into elementary frequency components [4].
The most common DCT definition of a 1-D sequence of
length N is

978-1-4673-0210-4/12/$31.00 2012 IEEE

B. Arnold Scrambling
The KK binary watermark image W is transformed into
W' by Arnold transformation to lower the autocorrelation
coefficient of image and then the confidentiality of watermark
is strengthened [6]. Arnold transformation is periodic and
when it is iterated some times the original signal will be
obtained. The Arnold transformation is given by

TRANSFORMS USED

The first transform coefficient is, referred to as DC

Coefficient, the average value of the sample sequence. The
other transform coefficients are called the AC Coefficients [5].

INTRODUCTION

The rapid development of the internet and the digital

information revolution cased significant changes in the global
society, ranging from the influence on the world economy to
the way people nowadays communicate [1]. Digitizing of
multimedia data has enabled reliable, faster and efficient
storage, transfer and processing of digital data. It also leads to
the consequence of illegal production and redistribution of
digital content. Digital watermarking is identified as a partial
solution which allows content creator to embed hidden data
such as author or copyright information into multimedia
content [2]. In general, all audio watermarking algorithms
work by exploiting the perceptual property of Human
Auditory System (HAS) [3]. Section II describes the
Transforms used in the method, Section III focuses on
embedding procedure, extraction algorithm is explained
Section IV, experimental results and conclusion are included
in sections V and VI respectively.
II.

(3)

(1)

530

2012 International Conference on Computing, Electronics and Electrical Technologies [ICCEET]

V.

Step 6: The overall watermarked audio signal is obtained by

summing all watermarked frames.

EXPERIMENTAL RESULTS

In the experiments a gray image of size 256x256 is used as

watermark shown in Fig.3b & original audio signal in Fig.3a.

Watermark Image (W)

Original audio signal
0.8

Arnold & DCT

Transformation

0.6

Original Audio Signal

f(x)

0.4
0.2
0
-0.2

Watermark
Embedding

Sampling &DCT

-0.4
-0.6
-0.8

8
4

IDCT

x 10

Watermarked Audio
Signal

Figure3a. Original Audio Signal

Figure1. Watermark Embedding

Watermark Image

IV.

WATERMARK EXTRACTON ALGORITHM

The process of watermark extraction is shown in Fig.2.

Watermarked Audio

Sampling & DCT

Transformation

Original Audio Signal

f(x)
Figure3b. Watermark Image

Coefficients
Subtraction

IDCT &

Arnold

DCT
Transformation
Experiments were conducted on four different classes of
signals such as speech, pop music, rock music and
instrumental as they differ in spectral properties. The
performance of the proposed technique is evaluated based on
MOS criteria, PSNR and Pearson Correlation.

Extracted Watermark
Image

Figure2. Watermark Extraction

A. Imperceptibility Test
Imperceptibility is related to the perceptual quality of the
embedded watermark image within the original audio signal.
To measure imperceptibility, we use Peak Signal-to-Noise
Ratio (SNR) as an objective measure, and a listening test as a
subjective measure. For subjective quality evaluation, a
listening test was performed with ten listeners to estimate the
Mean Opinion Score (MOS) grade of the watermarked signals
for four different signals [7]. Each listener was presented five
times with the pairs of original signal and the watermarked
signal and asked to report whether any difference could be
detected between the two signals. The MOS criteria are listed
in Table1. and the MOS values are tabulated in Table2.

Step 1: Sample the watermarked audio signal at a sampling

rate (fs) and partition the sampled file into frames each having
certain samples.
Step 2: Perform DCT transformation on watermarked audio
signal.
Step 3: Obtain the DCT coefficients of the original audio
signal.
Step 4: Subtract the coefficients of original signal from the
watermarked signal coefficients.
Step 5: Apply IDCT to the values obtained in Step 4.
Step 6: Perform Arnold operation to the values obtained in
Step 5.

531

2012 International Conference on Computing, Electronics and Electrical Technologies [ICCEET]

TABLE I. MOS CRITERIA

Score
5
4
3
2
1

VI.

A digital audio watermarking using Arnold and Discrete

Cosine Transform is proposed. The watermark image is
scrambled using Arnold and then embedded into DCT
coefficients of the original audio signal. The experimental
results show that the algorithm has good imperceptibility and
can resist some attacks such as resampling, noise, Gaussian
noise and requantization. But the results are poor in the case of
cropping attack.

Watermark Imperceptibility
Imperceptibility
Perceptibility but not annoying
Slightly annoying
Annoying
Very annoying
TABLE II. MOS VALUES

Class of Audio Signal

SPEECH
POP
ROCK
INSTRUMENTAL

REFERENCES

MOS
5.00
5.00
4.95
4.95

[1] Nedeljko, Cvejic Algorithms for Audio Watermarking and

SteganoGraphy,ISBN 951-42-7384-2, OULU University press.
[2] L.T.Bruton, J.D.Gordy, Performance Eval uation of Digital Audio
Watermarking Algor ithms,Proceedings of the 43rd IEEE Midwest
Symposium, Michigan, Volume 1, pp.456-459, 8-11 August 2000.

PSNR

[3] D. J. M. Robinson, M. J. Hawksford, Time-domain auditory model for

the assessment of high-quality coded audio, presented at the 107th
Convention of the Audio Engineering Society, New York, (1999 Sep.).

PSNR [8] is defined using the following equation:

PSNR 10log
(5)
In the above equation (5), X is the maximum fluctuation in the
images. MSE is the mean square error and defined as in
equation (6)

[4] Rao K and P.Yip, Discrete Cosine Transform algorithms, advantages,

applications. Academic Press, USA, 1990.
[5] The Discrete Cosine Transform(DCT): Theory and Application, Syed Ali
Khayam, De partment of Electrical & Computer Engineering, Michigan
StateUniversity, March 10th, 2003.

MSE=
(6)
*
Where I and I are the watermark image and the extracted
watermark image respectively.

[6] ZHAO Rui-mei, LIAN Hua ,PANG Hua-wei ,HU Bo-ning A Blind
Watermarking Algorithm Based on DCT of the Second International
Symposium on Intelligent Information Technology Application, IEEE
Computer Society,2008.

TABLE III. PSNR & PCC VALUES

S
P
E
E
C
H

P
O
P

R
O
C
K

I
N
S
T
R
U

CONCLUSION

Attack Type

PSNR

PCC

Without Attack
Resample
Noise
Filter
Requantization
Gaussian
Cropping
Without Attack
Resample
Noise
Filter
Requantization
Gaussian
Cropping
Without Attack
Resample
Noise
Filter
Requantization
Gaussian
Cropping
Without Attack
Resample
Noise
Filter
Requantization
Gaussian
Cropping

69.2375
53.9771
18.2499
2.2545
20.9370
18.1585
-3.3220
69.2509
49.9336
18.2493
-3.7940
20.9147
18.1240
-3.3166
69.2509
51.4242
18.2415
-2.6341
20.9119
18.0983
-2.6262
69.2509
50.4785
18.2654
-2.5802
20.9167
18.1034
-5.5148

1
1
0.9743
0.1687
0.9338
0.8927
0.1144
1
0.999
0.9743
0.0853
0.9337
0.8915
0.1100
1
0.9999
0.9743
0,0929
0.9336
0.8912
0.1240
1
0.999
0.9742
0.0976
0.9337
0.8909
0.0892

[7] Beerends J.and Stemerdink J., A Perceptual Audio Quality Measurement

Based on a Psychoacoustic Sound Representation, Journal of thr Audio
Engineering Society, vol. 40, no.12, pp.963-978, 1992.
[8] Liu Ping Feng, Liang Bin Zheng, Peng Cao, A DWT-DCT Based Blind
Watermarking Algorithm for Copyright Protection, IEEE Transactions, 2010

532