08-11-2012, 04:53 PM
Effective Pseudonoise Sequence and Decoding Function for Imperceptibility and Robustness Enhancement in Time-Spread Echo-Based Audio Watermarking
Effective Pseudonoise Sequence and Decoding.pdf (Size: 684.13 KB / Downloads: 28)
INTRODUCTION
RECENT advances in communication networks, computers
and multimedia technology facilitate the efficient
distribution, reproduction, and manipulation of digital media,
without any quality degradation. This has yielded strong demand
for the protection of digital media concerning intellectual
property rights. For example, the music industry loses millions
of dollars every year due to unauthorized copying and distribution
of music and other sound signals. Digital watermarking is
a promising technique to tackle this problem [1]–[7].
For the echo-hiding watermarking schemes in [6], [7], and
[21]–[23], security is a major problem since anyone can detect
the embedded echo delays and correspondingly the watermarks
using cepstrum analysis. This makes them vulnerable to malicious
attacks. To cope with this problem, Chen and Wu employ
the frequency hopping technique to transform, scramble, and
segment audio signal to increase security [28]. Each audio segment
has multiple fractions from different frequency bands of
the host audio signal. The manner of combining fractions is realized
by a pseudonoise (PN) sequence as a secret key.Without
knowing the PN sequence, it is difficult to extract the watermarks
embedded in the audio signal. In [1], Ko et al. propose a
simpler time-spread echo method to enhance security, where a
PN sequence is used to spread echoes. A challenging task for
the time-spread echo method is the tradeoff between robustness
and imperceptibility. More recently, a novel echo-hiding
watermarking method has been reported in the literature [3].
It is built upon the idea of time-spread echo hiding but aimed
at improving the watermark detection rate. This is achieved by
shifting the host audio signal effect in the decoder to the watermarked
signal at the encoding stage. However, this reduces
the perceptual quality of the watermarked signal. Furthermore,
shifting the host audio signal effect in the decoder to the watermarked
signal significantly increases the encoding time. Therefore,
this method is costly in computation.
EVALUATION
In the evaluation, we jointly use the proposed PN sequence
and the decoding function for watermarking. Simulations
are conducted to demonstrate the imperceptibility and
robustness of the new watermarking method, in comparison
with the echo-hiding-based methods in [1] and [3]. Listening
test is also used to assess the perceptual quality of our method.
We use 250 audio clips belonging to five different audio groups
as host audio signals, which are as follows:2
• Group 1: 50 clips containing Indian and Sri Lankan classical
music;
• Group 2: 50 clips containing Russian, Indian, and western
folk music;
• Group 3: 50 clips containing English and Sri Lankan
country music;
• Group 4: 50 clips containing male and female speeches;
• Group 5: 50 clips containing mixture of bird, animal, wind,
sea wave, and rain sounds.
All these audio clips are mono-channel and have a duration of 20
s. They are sampled at the rate of 44.1 kHz, quantized with 16
bits, and then segmented. Each segment contains 44 100 samples,
all of which are utilized in computing discrete Fourier
transform (DFT). For each audio clip, a different PN sequence
is employed to embed watermarks. The length of the PN sequences
and is chosen to be 1023.
CONCLUSION
In this paper, we propose an effective PN sequence for timespread
echo-based audio watermarking. The proposed sequence
yields an echo kernel whose magnitude response has smaller
magnitudes in the perceptually significant region. Furthermore,
its correlation function has more large peaks than that of the existing
PN sequence, from which a modified decoding function
is derived. As a result, the new audio watermarking method that
jointly employs the proposed PN sequence and decoding function
can significantly enhance the perceptual quality of the watermarked
signal and the robustness of time-spread echo-based
audio watermarking. The superior performance of our method is
demonstrated by simulation examples and listening test, in comparison
with the audio watermarking methods in [1] and [3].