Más contenido relacionado
La actualidad más candente (19)
Similar a 40120130406010 2-3-4-5-6 (20)
Más de IAEME Publication (20)
40120130406010 2-3-4-5-6
- 1. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
INTERNATIONAL JOURNAL OF ELECTRONICS AND
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
COMMUNICATION ENGINEERING & TECHNOLOGY (IJECET)
ISSN 0976 – 6464(Print)
ISSN 0976 – 6472(Online)
Volume 4, Issue 6, November - December, 2013, pp. 71-84
© IAEME: www.iaeme.com/ijecet.asp
Journal Impact Factor (2013): 5.8896 (Calculated by GISI)
www.jifactor.com
IJECET
©IAEME
A HIGH CAPACITY DIGITAL AUDIO WATERMARKING USING
DISCRETE WAVELET AND COSINE TRANSFORM
G. B. Khatri,
Dr. D. S. Chaudhari
Department of Electronics and Telecommunication,
Government College of Engineering,
Amravati, India.
ABSTRACT
Piracy and copyright protection is a major issue among the music owners and publishers.
Major improvements in digital and internet technologies have made the illegal production, storage
and distribution of the digital multimedia files very easy. As a remedy digital audio watermarking is
nowadays gaining popularity among the original audio content producers. This technique embeds the
owner information as a copyright material in the digital audio as a proof of ownership. This paper
introduces the watermarking, its applications and techniques developed. An algorithm is proposed
based on wavelet transform which has efficient computational load and also on cosine transform for
increasing the watermark embedding capacity. Algorithm is attacked with different signal processing
attacks and then evaluated on the basis of watermarking parameters like imperceptibility, robustness,
and embedding capacity.
Index Terms: Watermark, Audio, DRM, DFT, DCT, DWT and DSSS.
I. INTRODUCTION
The chain of production, sales and copyright has been drastically disturbed due to invention
and widespread use of digital media and internet. Virtuality of digital media affects the economic
behind the production and distribution of media and the copyright of the original media. The piracy
of non-physical form of media is due to their digitalization; this gets exaggerated due to negligible
cost of reproduction and their ability to be digitally delivered.
The production, storage, distribution of digital multimedia data is very easy over the internet.
Hence it creates the problem of protecting the intellectual copyrights and the ownership of the media.
The copyrighted digital multimedia data is pirated without notification to the owner.
Digital watermarking is a technique to embed the owner information as a copyright material
in the digital data as a proof of ownership. Watermark as the name suggest is as transparent as water
71
- 2. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
when watermark data is embedded in the original audio. Thus a watermark as a signature can be
embedded in the digital multimedia files as a proof of ownership [1, 2].
Digital watermarking can be applied to image, audio or video and the watermark data can be
an image, audio and text. Generally there is less attention towards audio watermarking because HAS
(human auditory system) is more sensitive than HVS (human visual system) and human ears can
easily detect the presence of the watermark as low as one part in ten million [3].
This paper presents an overview on applications and techniques of digital audio
watermarking. Section II describes the digital audio water marking, its applications and previous
watermarking techniques while in section III we describe in details the watermark embedding and
extraction procedures of the proposed algorithm; while the performance of the algorithm and
simulation results with respect to inaudibility, robustness and payload capacity is presented in
Section IV. Section V concludes the paper with some remarks.
II. DIGITAL AUDIO WATERMARKING: APPLICATIONS AND TECHNIQUES
Digital audio watermarking is a technique of embedding watermark data such as image, audio
and text in the original audio stream to create copyrighted watermarked audio. Audio watermarking
application areas include such as vendor identification, evidence of proprietorship, validation of
genuineness, copy protection, etc. Each application puts desirable feature necessity on the
watermarking technique. Hence the watermarking technique to be used depends on the area of
application [4]. Thus a variety of applications are discussed below,
1) Vendor Identification
Text form of copyright notices occurs on the packaging of copyrighted materials. This type of
protection does not prove sufficient as it would be easily removed. Digital audio watermarking can
be used to embed copyright notice in the audio signal itself. As notice forms an integrated part of
audio one can determine the vendor of the copyrighted audio.
2) Evidence of Proprietorship
One can prove its proprietorship in the case of copyright dispute. The original owner can
prove its proprietorship by extracting the watermark copyright information from the watermarked
audio, in the case when another person tries to sell the copyrighted material on behalf of his name by
pirating it.
3) Validation of Genuineness
The copyrighted audio is genuine or not, can be proved very easily by the use of
watermarking. A signature or copyright watermark is embedded in the audio thus anyone trying to
modify the watermarked audio, modification also applies to the watermark because watermark forms
integral part of audio. Hence one can prove genuineness of copyrighted audio by extracting exact
copy of the watermark.
4) Copy Protection
Above mentioned applications do not put restrictions on the illegal copying. Owner can
restrict the illegal copying or the numbers of copies permitted by the use of special watermarking
algorithms. One such method is modified patchwork algorithm (MPA) developed by Yeo and
Kim [5].
Audio watermarking techniques can be grouped as; time-domain, frequency (transform)
domain, spread-spectrum, patchwork [1, 5].
72
- 3. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
1) Time-domain techniques
It includes the least significant bit (LSB) substitution, echo hiding and quantization
techniques. The time-domain audio watermarking is relatively easy to implement, and requires few
computing resources, however, it is weak against signal processing attacks such as compression and
filtering. This type of techniques suffers from problems like low watermark data embedding
capacity, easily detectable by the attacker, easy to decode watermark from cover audio.
LSB technique is the simplest technique in which watermark details embedded in the least
significant bits of the audio sample values. As the information contained in the LSB is less, it is
replaced by the data of the watermark without producing the noticeable effect in the cover audio
signal. A maximum of 3 watermark bits per 16 bits of audio sample is allowed for imperceptibility.
Cvejic N. and Seppanen T. have tried to increase the capacity from 3bits/sample to 4bits/sample
without degrading the watermarked audio signal to noise ratio by using a three step technique. In this
degradation in SNR of the watermarked audio is minimised by using minimum error replacement
and error diffusion steps.
Echo-hiding watermarking embeds information into the original discrete audio signal by
introducing a repeated version of an original sample of the audio signal with some delay and decay
rate so as to make it undetectable [6]. Digital data is embedded by using four main parameters of
echo: initial value, decay rate and different offset for 1 and 0. The offset is made so small such that
the human ear cannot detect the presence of echo. The watermark data embedding rate is given as 16
bps (bits per second), while it can vary in the range 2–64 bps and it depends on the sampling rate and
the signal type to be echoed.
In the technique of quantization original sample of audio is replaced with the modified audio
sample. The modified audio sample is defined as below,
ݕൌቐ
ݍሺܣ ,ݔሻ ସ , ݂݅ ݀ܽ1 ݏ݅ ݐܾ݅ ܽݐ
ݍሺܣ ,ݔሻ െ ସ , ݂݅ ݀ܽ0 ݏ݅ ݐܾ݅ ܽݐ
(1)
Where q (.) is quantization function and is quantization step. The quantization function is given as,
ݍሺܣ ,ݔሻ ൌ ሾܣ⁄ ݔሿ
(2)
Where [x/A] is rounded to nearest integer. Thus in a single sample of audio signal one can embed
only one bit of watermark. Hence a blind detection can be applied for watermark data extraction.
Extraction can be done by following equation,
1 , ݂݅
0 ൏ ݕെ ݍሺܣ ,ݔሻ ൏ 4/ܣ
ܾൌ൜
(3)
0 , ݂݅ െ 4/ܣ൏ ݕെ ݍሺܣ ,ݔሻ ൏ 0
This technique is simple and easy to implement and is robust to noise as long as the noise margin is
below A/4. While the technique is easy but the watermark embedding capacity is very less.
2) Frequency domain techniques
Frequency domain audio watermarking techniques generally include transforms like discrete
Fourier transform (DFT), the discrete cosine transform (DCT), and the discrete wavelet transform
(DWT). It takes the advantage of masking of different tones of human auditory system (HAS) for
effective watermarking. Discrete Fourier transform decomposes the signal into its fundamental and
harmonically related sinusoidal frequencies. The human ears sensitivity declines after the peak
sensitivity around 1 kHz. Magnitude response coefficients are replaced by the watermark data in the
frequency range of 2.4 – 6.4 kHz [7]. Also the human ears are insensitive to the absolute phase of the
audio frequency; hence the phase difference between the phase signal coefficient and phase reference
coefficient is used to modify the phase signal coefficient [8].
Discrete cosine transform is similar to the discrete Fourier transform except that its
coefficients are real valued. Properties of DCT such as high compaction of signal energy in
transform domain, highly decorrelated coefficients are used to embed data in the transform domain.
73
- 4. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
Discrete wavelet transform is nowadays gaining popularity because it can decompose the
signal in time and frequency at the same time while keeping the calculations to obtain DWT
coefficients small as compared to DFT and DCT. Several advantages of applying DWT on audio
signal are given by wu and huang such as a) It is able to localize the audio in time-frequency both
with multi-resolution property, b) variable decomposition levels are available, c) less number of
operations than DFT and DCT [9]. If there are N samples in the audio then number of operations in
DFT, DCT and DWT are O (N·log2(N)), O(N·log2(N)) and O(L·N) respectively, where L is the
length of wavelet filter. A data payload capacity of 172 bps is achieved by embedding the selfsynchronised watermark data in the wavelet domain without degrading the SNR too much[9].
3) Spread-spectrum technique
It involves embedding of watermark in the original audio signal by spreading it over the
bandwidth of audio signal [6]. This technique utilizes DSSS (Direct Sequence Spread Spectrum)
method of spread spectrum.
In DSSS PN-sequence in used to spread the watermark data in the whole frequency range of
audio and then added to the audio signal by proper attenuation, so that the watermark data is treated
as additive random noise. Same sequence is again used to extract the watermark data by performing
correlation between watermarked audio and PN-sequence.
4) Patchwork technique
This technique was first proposed for image watermarking and it is a pseudorandom
statistical approach [6]. The idea is to select two subsets (patch) of the cover signal and in order to
embed the watermark, the sample values of these two subsets are moved in opposite directions by a
constant valued, which defines the watermark strength and watermark bit. The assumption in this
method is that the difference of the means of the two patches is zero for the original cover signal and
is nonzero for the watermarked cover signal. As two subsets (patch) are used this technique can
extract the watermark without the original cover signal.
Yeo and Kim have proposed a modification on the patchwork technique [5]. They have
embedded the watermark data bit in the two subsets taken from DCT domain of original audio
signal. The use of transform domain for embedding watermark makes the technique more robust
against signal processing attacks such as down-sampling, equalization, compression, filtering.
III. PROPOSED ALGORITHM
As observed from the above techniques that transform domain are more secure than time
domain. The models proposed utilize the discrete wavelet transform for speedily and efficiently
transforming audio in time-frequency domain, while using discrete cosine transform to decor relate
and compress watermark image.
Transformed watermark image coefficients must be normalised and multiplied with an
attenuation constant before embedding. Attenuating the coefficients helps to keep noise level low in
the audio signal. Since watermark image is compressed using discrete cosine transform less number
of transform coefficients are used for embedding and this improves the signal to noise ratio and also
the watermark data embedding capacity.
Discrete wavelet transform is used to transform the audio signals into frequency sub bands.
These sub bands are called as approximate or detail frequency sub bands. Any one of this frequency
band can be used to embed watermark.
This algorithm is based on emerging wavelet transform and cosine transform for embedding
and extracting watermark in audio signal, hence it will possess characteristics such as high signal-to-
74
- 5. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
noise ratio (SNR) of watermarked audio, high data payload capacity i.e. number of watermark bits
per second of audio signal, low computational complexity.
1) Embedding algorithm
1. The original audio signal is decomposed at level N with the help of 1D-Discrete Wavelet
Transform (1D-DWT) and accordingly we get wavelet coefficients as,
ܿ ൌ ܹܶܦሺݕሺݐሻሻ
ܿ ൌ ܿܣே , ܿ ܦே , ܿܦேିଵ , … … … … , ܿܦଵ
ܿܣே ൌ ܰ ݈݁ݒ݈݁ ݐܽ ݏݐ݂݂݊݁݅ܿ݅݁ܥ ݊݅ݐܽ݉݅ݔݎܣ
ܿ ܦே , ܿ ܦேିଵ , … … … … , ܿܦଵ ൌ 1 ݈݁ݒ݈݁ ݐ ܰ ݈݁ݒ݈݁ ݉ݎ݂ ݏݐ݂݂݊݁݅ܿ݅݁ܥ ݈݅ܽݐ݁ܦ
Figure 1 Wavelet Decomposition
Figure 2 Watermark Embedding Model
75
(4)
(5)
(6)
(7)
- 6. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
2. Watermark is applied Discrete Cosine Transform (DCT) so as to concentrate watermark energy in
one corner. This will form a set of DCT coefficients in the form of matrix having the dimensions
as that of the watermark image.
Figure 3 Watermark Image
Figure 4 DCT Transformed Watermark
3. Matrix of DCT transformed watermark is converted into row stream W by zigzag scanning
method.
4. Multiply the row stream of watermark by alpha blending constant (α) for controlling the
brightness of extracted watermark and for achieving high Signal to Noise Ratio (SNR) for the
watermarked audio signal.
(8)
ܹ′ ൌ ߙ ൈ ܹ
5. Truncate the row of DCT coefficients to a limited number of coefficients as provided by the user.
This will form an instance of DCT coefficients of the watermark.
ܹ ′′ ൌ ݁ݐܽܿ݊ݑݎݐሺܹ ′ ሻ
(9)
6. Append the instances of watermark in series and then add to one of the sub-bands of the wavelet
decomposed audio signal.
ܹ ′′′ ൌ ݉ݏ݁݅ݎ݁ݏ ݊݅ ′′ ܹ ݏ݁ܿ݊ܽݐݏ݊݅ ݈݁݅ݐ݈ݑ
(10)
′′′
݈݁݊݃ ܾݑݏ ݀݁ݐ݈ܿ݁݁ݏ ݂ ݄ݐെ ܾܽ݊݀ ݈ܹ݁݊݃ ݂ ݄ݐ
(11)
ଵ′
ଵ
ܿ ܦൌ ܿ ܦ ܹ
(12)
7. Apply 1D-Inverse Discrete Wavelet Transform (1D-IDWT) of level N to watermark embedded
audio signal so as to create the watermarked audio signal.
′
ܿ ′ ൌ ܿܣே , ܿܦே , ܿܦேିଵ , … … … … , ܿܦଵ
(13)
′
′
ݕሺݐሻ ൌ ܹܶܦܫሺܿ ሻ
(14)
2) Extracting algorithm
1. The original audio signal and watermarked audio signal is decomposed at level N with the help of
1D-Discrete Wavelet Transform (1D-DWT).
ܿ ൌ ܹܶܦሺݕሺݐሻሻ
(15)
ܿ ൌ ܿܣே , ܿ ܦே , ܿ ܦேିଵ , … … … … , ܿܦଵ
(16)
′
′
(17)
ܿ ൌ ܹܶܦሺݕሺݐሻ ሻ
′
ே
ே
ேିଵ
ଵ′
ܿ ൌ ܿܦܿ , ܦܿ , ܣ
, … … … … , ܿܦ
(18)
2. Subtract the sub-band of original audio signal from sub-band of watermarked audio signal
containing watermark, so as to recover the instances of DCT coefficients of watermark.
′
ܹ ൌ ܿܦଵ െ ܿܦଵ
(19)
76
- 7. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
Figure 5 Watermark Extraction Model
3. Extract any one of the instance from the series of instances recovered from watermarked audio
signal.
4. Divide the row stream of extracted watermark by alpha blending constant (α).
5. Convert row stream of extracted watermark in to matrix form by inverse zigzag scanning method.
6. The matrix obtained is processed by Inverse Discrete Cosine Transform (IDCT) to recover
watermark in its original form.
IV. RESULTS AND PERFORMANCE ANALYSIS
This research work uses a mono channel audio signal with .WAV format, 8 kHz sampling
frequency, 16 bit depth, and 13.8114 sec long as a cover signal. While for the watermark a grayscale
image with dimension 252×298 and bit depth of 8 bit in .TIF format is chosen. All algorithms,
including proposed technique, are implemented on Windows PC having Intel Core-i3 2.93 GHz
processor, 2GB RAM and are simulated using MATLAB2009.
Results are based on the parameters like Imperceptibility, Robustness and Payload
Capacity. The robustness performance of the algorithm is evaluated under different signal processing
77
- 8. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
attacks like re-quantization, re-sampling, low-pass filtering, high-pass filtering, AWGN and
cropping. Following paragraph describes the performance metrics and then the results are presented.
1) Imperceptibility
Imperceptibility is related to the perceptual quality of the embedded watermark data within
the original audio signal. To measure imperceptibility, we use Signal to Noise Ratio (SNR) as
quantitative measure, and a listening test as a qualitative measure.
Signal to Noise Ratio (SNR) is a statistical difference metric which is used to measure the similarity
between the original audio signal and the watermarked audio signal.
ܴܵܰሺ݀ܤሻ ൌ 10 ݈݃ଵ
∑ሺ ሻమ
∑൫ ି ′ ൯
మ
(20)
Where, Anis original audio signal and A’n is watermarked audio signal.
A listening test was performed to estimate the qualitative grade of the watermarked signals. These
grades are based on the criteria of Mean Opinion Score (MOS) which describes the perception
quality.
Table 1 Mean Opinion Score (MOS) Grade
MOS Grade
Description
5
Imperceptible
4
Perceptible, but not annoying
3
Slightly annoying
2
Annoying
1
Very annoying
2) Robustness
Watermarked audio signals can undergo common signal processing attacks such as filtering,
compression, cropping, time scale modification and many others. These attacks happen either
intentionally or unintentionally. Although these attacks may not affect the perceived quality of the
host signal, they can corrupt the watermark embedded within the signal. To evaluate robustness of
the proposed algorithm, we implemented a set of attacks that commonly affect audio signals.
After extracting watermark from corrupted audio signals they are compared with original watermark
for their similarity. This similarity is calculated by finding the correlation factor (ρ) between the two
of them.
ߩሺ ′ ݓ ,ݓሻ ൌ
′
∑ಿ ሺ௪ ൈ ௪ ሻ
సభ
′
ට∑ಿ ሺ௪ ሻమ ൈ ට∑ಿ ൫௪ ൯
సభ
సభ
మ
(21)
Where w and w’ are original watermark and extracted watermark, respectively.
3) Payload capacity
The payload capacity C of the algorithm to embed the amount of watermark bits per second
of original audio signal is calculated by the following formula;
்௧ ௨ ௧௦
ܥൌ
(22)
ை ௨ௗ ௗ௨௧
ܥൌ
ே௨ ௫௦ ൈ௧ ௗ௧ ௫
௨ௗ ௗ௨௧
ሺܾݏሻ
(23)
Generally due to use of Discrete Cosine Transform (DCT) the payload capacity C varies
drastically. This happens because only a length of DCT coefficients is used for embedding the
watermark. This length depends on the sub-band of DWT transformed original audio used for
watermark embedding.
78
- 9. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
In the simulation for the analysis and synthesis of the audio signal is done using dB1 wavelet
with level 1 decomposition. The alpha constant (α) is set to 0.1 and the first 10000 DCT coefficients
of DCT transformed image is used to embed watermark in the first level detail sub band of DWT
decomposed audio. Table 2 shows the results based on imperceptibility, robustness, payload capacity
under different attacks.
Table 2 Results of proposed algorithm under various attacking conditions
Payload
Imperceptibility
Robustness
Capacity
Attacks
Correlation Factor
C (bps)
SNR (dB)
MOS
(ρ)
Un-attacked
58.1767
5
0.9982
43498.0586
Low Pass Filter
58.1767
5
0.8426
43498.0586
High Pass Filter
58.1767
5
0.9968
43498.0586
Resampling (4000)
58.1767
5
0.9148
43498.0586
Resampling (16000)
58.1767
5
0.7720
43498.0586
Requantization (8-bit)
58.1767
5
0.6275
43498.0586
Requantization (24-bit)
58.1767
5
0.9982
43498.0586
Cropping 25% samples
58.1767
5
0.9982
43498.0586
Amplification (20%)
58.1767
5
0.9982
43498.0586
Amplification (-20%)
58.1767
5
0.9982
43498.0586
AWGN (90dB)
58.1767
5
0.9934
43498.0586
High value of SNR shows the similarity between original and watermarked audio, while the
Figure 6 shows it graphically.
Figure 6 Original and Watermarked Audio
79
- 10. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
In Figure 6 one can observe that five instances of watermark in the form of DCT coefficients
are embedded in series in the audio. The peaks in the error audio show the first sample location of
the DCT coefficient.
Extracted watermark is shown in the Figure 7 below for various signal processing attacks.
(a)ρ = 0.8426
(f) ρ =0.9982
(b)ρ =0.9968
(g) ρ =0.9982
(c)ρ = 0.9148
(d) ρ = 0.7720
(h) ρ =0.9982
(i) ρ =0.9982
(e)ρ =0.6275
(j)ρ =0.9934
Figure 7 Extracted Watermark and their Correlation Factor (ρ) Under Different Attacks
(a)Low Pass Filter, (b)High Pass Filter, (c)Downsampling, (d)Upsampling, (e)Requantizaton to 8-bit,
(f)Requantizaton to 24-bit,(g)Cropping 25%, (h)Amplification 20%, (i)Amplification -20%,
(j)Gaussian Noise
In Figure 7 it is observed that the watermark is robust to most of the attacks except
resampling (Down and up) and requantization (Down). The similarity i.e. correlation factor (ρ)
between the original and watermarked audio signal is also shown which is near to 1. The high value
of ρ indicates that the system is robust to most of the signal processing attacks.
The performance depends on proper selection of α; Graph1 provides the effect of alpha
constant (α) on the signal to noise ratio and on the correlation factor (ρ) of the watermarked signal. A
low SNR value yield high robustness but the quality of the watermarked signal is degraded as shown
in graph using high α; and lower value of α gives high SNR which make the watermark
imperceptible, however robustness is decreased. Hence, α value is chosen in a way that the SNR of
the watermarked signal is between 40 dB and 70 dB for efficient performance. The results shown in
Graph 1 and Graph 2 are for no signal processing attack on watermarked audio signal.
80
- 11. SNR (dB)
International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
120
110
100
90
80
70
60
50
40
30
20
10
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Alpha Constant (α)
Correlation Factor ρ
Graph 1 Effect of Alpha Constant (α) on Signal to Noise Ratio (SNR)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Alpha Constant (α)
Graph 2 Effect of Alpha Constant (α) on Correlation Factor (ρ)
The most suitable value of α from the Graph 1 and Graph 2 is found to be 0.1 and this value
of α gives SNR value above 55dB and value of ρ close to 1. The high similarity between original and
extracted watermark is due to higher values of ρ.
Performance of the algorithm depends on the number of DCT coefficients embedded. Graph3
provides the effect of number of DCT coefficients embedded on the signal to noise ratio and on the
correlation factor (ρ) of the watermarked signal. Embedding less number of DCT coefficients
increases the number of instance in the watermarked audio which increases the robustness but the
quality of the watermarked signal is degraded as shown in graph; and if more number of DCT
coefficients are embedded SNR is increased which make the watermark imperceptible, however
robustness is decreased. Hence, number of DCT coefficients is chosen in a way that the SNR of the
watermarked signal is between 40 dB and 70 dB for efficient performance without degrading the
quality of extracted watermark i.e. increase in robustness. Considering the same conditions for the
experiment following graphs are obtained by varying the number of DCT coefficients to be
embedded.
81
- 12. SNR (dB)
International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
70
65
60
55
50
45
40
35
30
25
20
15
10
5
0
0
2500
5000
7500
10000
12500
15000
17500
20000
Number of DCT coefficients embedded
Graph 3 Effect of Number of DCT Coefficients Embedded on Signal to Noise Ratio (SNR)
Correlation Factor ρ
1
0.9
0.8
0.7
0.6
0.5
0
2500
5000
7500
10000 12500 15000 17500 20000
Number of DCT coefficients embedded
Graph 4 Effect of Number of DCT Coefficients Embedded on Correlation Factor (ρ)
From the results and analysis it can be stated that the imperceptibility of the system largely
depends on the values of the alpha constant (α) and number of DCT coefficients of watermark to be
embedded. The imperceptibility of the system is optimised when the value of α is 0.1. The length of
the level-1 detail subband is maximum and this is the reason for choosing the level-1 detail subband
for embedding audio signal. Imperceptibility also depends on the value of the number of DCT
coefficients embedded; hence from the results it is shown that 10000 is the optimised value.
V. CONCLUSION
In this dissertation a new algorithm has been proposed taking features of DWT and DCT;
where the DWT is the developing area in signal processing theories. Proposed algorithm is based on
signal energy compaction in DCT domain while DWT as a faster approach to convert the signal into
transform domain.
82
- 13. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
The performance of the algorithm is provided by evaluating the performance parameters such
as signal to noise ratio SNR, correlation factor (ρ), and MOS. Capacity as a performance parameter is
the goal of this developed algorithm and has been achieved by using DCT which is most efficient
and popular for compacting energy of the signal onto very few coefficients. There is no effect on the
payload capacity due to the number of coefficients to be embedded because it is calculated on the
basis of original watermark bits. If less number of DCT coefficients are embedded the extracted
image gets somewhat blurred and hence we never gets the correlation factor (ρ) equals to 1.
Performance of the algorithm is evaluated by applying the signal processing attacks. The algorithm
proposed by using DWT and DCT transformation is not robust against resampling. In addition, the
effects of the alpha blending constant (α) and the number of DCT coefficients to be embedded on the
SNR and on correlation factor (ρ) is shown and analysed.
The audio watermarking is relatively new and has wide scope for research. This research can
be continued to colour images from the limitation of embedding only grayscale image. Also the
technique can be implemented on live signals rather than a fixed signal. Some of the real time audio
signals include speech and music. Further research can be carried on embedding watermark in video
sequences i.e. movies or surveillance.
REFERENCES
[1] Wikipedia.(2013).DigitalWatermarking.Available:
http://en.wikipedia.org/ wiki/Digital_watermarking.
[2] A. Al-Haj, A. Mohammad, and L. Bata, “DWT based Audio Watermarking,” The International
Arab Journal of Information Technology, vol. 8, no. 3, 2011, pp. 326–333.
[3] Gruhl D., Lu A., Bender W., “Echo Hiding,” in Proceedings of the International Conference on
Info Hiding, pp. 295–315, 1996.
[4] Cox I., Miller M., Bloom J., “Watermarking applications and their properties,” International
Conference on Information Technology, Las Vegas, pp. 1–5, 2000.
[5] Yeo I. and Kim H., “Modified Patchwork Algorithm: A Novel Audio Watermarking scheme,”
IEEE Transactions on Speech and Audio Processing, vol. 11, no. 4, pp. 381–386, 2003.
[6] Bender W., Gruhl D., Morimoto N., Lu A., “Techniques for data hiding,” IBM Systems Journal,
vol. 35, no. 3 and 4, pp. 313–336, 1996.
[7] Tilki J. F., Beex A. A., “Encoding a Hidden Digital Signature onto an Audio Signal Using
Psychoacoustic Masking, ”7th International Conference on Signal Processing Applications &
Technology, Boston MA, pp. 476–480, 1996.
[8] Tilki J. F., Beex A. A., “Encoding a Hidden Auxiliary Channel onto a Digital Audio Signal using
Psychoacoustic Masking,” IEEE Southeastcon, Blacksburg, VA, pp. 331–333, 1997.
[9] Wu S., Huang J., “Efficiently Self-Synchronized Audio Watermarking for Assured Audio Data
Transmission,” IEEE Transactions on Broadcasting, vol. 51, no. 1, pp. 69–75, 2005.
[10] Khatri G. B., Chaudhari D. S., “Digital Audio Watermarking Applications and Techniques,”
International Journal of Electronics and Communication Engineering and Technology (IJECET),
IAEME, Vol. 4, Issue 2, pp. 109-115, March–April 2013.
[11] Mr. N. R. Bamane, Dr. Mrs. S. B. Patil, Prof. B. S. Patil and Prof. R. K. Undegaonkar, “Hybrid
Video Watermarking Technique by using DWT & PCA”, International Journal of Electronics and
Communication Engineering and Technology (IJECET), IAEME, Vol. 4, Issue 2, pp. 172 - 179,
March–April 2013
[12] Shefaly Sharma and Jagpreet Kaur, “A Robust and Secure Visible Watermarking Scheme Based
On Multi Wavelet Technique With Modifird Fast Haar Transform [MFHWT]” International
Journal of Advanced Research in Engineering & Technology (IJARET), IAEME, Vol. 4, Issue 2,
pp. 240 - 247, ISSN Print: 0976-6480, ISSN Online: 0976-6499, 2013.
83
- 14. International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 –
6464(Print), ISSN 0976 – 6472(Online) Volume 4, Issue 6, November - December (2013), © IAEME
AUTHORS DETAILS
Govind B. Khatris pursuing M.Tech. in Electronic System and
Communication from Government College of Engineering, Amravati
(Autonomous Institute of M.S.). He received his B.E. in Electronics and
Telecommunication from H.V.P.M.’s College of Engineering & Technology in
2011. He also received Diploma in Electronics and Communication from
Government Polytechnic, Amravati (Autonomous Institute of M.S.) in year
2008. His area of research includes Digital Signal Processing, Wavelet
Processing, Digital Watermarking, and HDL with FPGA.
Dr. Devendra S. Chaudhari obtained BE, ME, from Marathwada
University, Aurangabad and PhD from Indian Institute of Technology,
Bombay, Mumbai. He has been engaged in teaching, research for period of
about 25 years and worked on DST-SERC sponsored Fast Track Project for
Young Scientists. He has worked as Head Electronics and Telecommunication,
Instrumentation, Electrical, Research and in charge Principal at Government
Engineering Colleges. Presently he is working as Head, Department of
Electronics and Telecommunication Engineering at Government College of Engineering, Amravati.
Dr. Chaudhari published research papers and presented papers in international conferences abroad at
Seattle, USA and Austria, Europe. He worked as Chairman / Expert Member on different committees
of All India Council for Technical Education, Directorate of Technical Education for Approval,
Graduation, Inspection, Variation of Intake of diploma and degree Engineering Institutions. As a
university recognized PhD research supervisor in Electronics and Computer Science Engineering he
has been supervising research work since 2001. One research scholar received PhD under his
supervision. He has worked as Chairman / Member on different university and college level
committees like Examination, Academic, Senate, Board of Studies, etc. he chaired one of the
Technical sessions of International Conference held at Nagpur. He is fellow of IE, IETE and life
member of ISTE, BMESI and member of IEEE (2007). He is recipient of Best Engineering College
Teacher Award of ISTE, New Delhi, Gold Medal Award of IETE, New Delhi, Engineering
Achievement Award of IE (I), Nashik. He has organized various Continuing Education Programmes
and delivered Expert Lectures on research at different places. He has also worked as ISTE Visiting
Professor and visiting faculty member at Asian Institute of Technology, Bangkok, Thailand. His
present research and teaching interests are in the field of Biomedical Engineering, Digital Signal
Processing and Analogue Integrated Circuits.
84