19
Audio Watermarking Under the guidance of Dr. Saroj Kumar Meher Presented by Suresh Chandra Martha EC200118335

Audio watermarking

Embed Size (px)

DESCRIPTION

 

Citation preview

Page 1: Audio watermarking

Audio Watermarking

Under the guidance of

Dr. Saroj Kumar Meher

Presented by

Suresh Chandra MarthaEC200118335

Page 2: Audio watermarking

INTRODUCTION

Digital Data can be shared by multiple users, distributed over network, and managed for long period time without any damage.

In contrast to these, the copyright protection problem arises, since unauthorized copying and distribution of digital data are simplified, too.

As a result, a technique called digital watermarking is introduced to protect the ownership of these contents.

Page 3: Audio watermarking

AUDIO WATERMARKING

Digital Audio Watermarking is a technology to hide information in an audio file without the information being audible to the listener, and without affecting in any way the audio quality of the original file.

What we can define

AIMS AT• Copy control of digital music• Ownership identification• Enforcement of usage policy • Fingerprinting

Page 4: Audio watermarking

AUDIO WATERMARKING CHARACTERISTICS

Transparency: watermark is not detectable

Robustness: survives digital processing (e.g., compression) and malicious analytical attacks.

Efficiency: low cost implementation using minimal resources

Independence: of both the type and format of multimedia data (e.g., music, speech, video, sample rates, coding scheme, ...)

Page 5: Audio watermarking

CLASSIFICATION OF AUDIO WATERMARKING

Temporal watermarking

Temporal watermarking hides watermarks directly into digital audio signals in the time domain.

Spectral water-marking

The spectral audio watermarking applies certain frequency transform, such as FFT, DCT, and DWT, etc, to the data block of the audio signal, and hides the watermark information Into the transformed data block.

Page 6: Audio watermarking

AUDIO WATERMARKING SCHEMES

DC level shifting

QMF bank band division

Frequency masking

Spread spectrum

Page 7: Audio watermarking

DC LEVEL SHIFTING

Here watermark is embedded by shifting the DC level of the audio signal.

An input signal is divided into frames of fixed length.

DC level of each frame =Mean of a frame-value in the frame

After that a binary watermark sequence generated randomly is introduced to the shifted signal. This is achieved by introducing a DC offset level to each frame in the signal according to the watermark bits.

If the watermark bit is 0, the signal is shifted downward.

If the watermark bit is 1, the signal is shifted upward. Cont..

Page 8: Audio watermarking

DC LEVEL SHIFTING

level0=-DCBiasMultiplier x FramePower

level1=+DCBiasMultiplier x FramePower

wherelevel0 ->DC level introduced when a watermark bit is 0

level1 -> the DC level introduced when a watermark bit is 1.

To extract a watermark, a watermarked signal is first divided into frames. Then, the mean of each frame is calculated. If the mean of a frame is positive, the corresponding watermark bit is 1. Else, the watermark bit is 0.

The original signal and watermark is not needed in the extraction process. Therefore, DC level shifting is a blind watermarking scheme

Page 9: Audio watermarking

Band Division Based on Quadrature Mirror Filter (QMF) Bank

Cont..

Page 10: Audio watermarking

Cont..

Page 11: Audio watermarking

Watermarked signal is filtered by the analysis filter bank and the low frequency watermarked band is extracted out.

The modified MDCT coefficients are calculated out and watermark can be determined by rounding the coefficients and observe whether coefficients are even or odd.

Watermark is 1 when coefficients are odd.

Watermark is 0 when coefficients are even.

This technique has a

High signal to noise ratio

The rate of detection or bit error rate is very high

WATERMARKED SIGNAL EXTRACTION USING QMF

Page 12: Audio watermarking

FREQUENCY MASKING Watermarking involves hiding a faint but audible sound under another

louder sound (original sound), so the faint sound becomes inaudible

and it acts as a watermark for the original sound.

Watermark is generated firstly by creating a pseudo-noise sequence

(PN sequence) from a shift register. The shift register consists of m flip-

flops with a maximum period of (2m) – 1.

PN sequence is filtered along with the filter that approximates the

frequency characteristics of the human auditory system (HAS).

Consequently, it is put through to the time domain for temporal

masking. Cont..

Page 13: Audio watermarking

The frequency masking algorithm is robust to coding, multiple watermarking, and resampling.

Different watermark can be embedded by this algorithm into the same signal and is easy to detect by the owner.

The detection scheme for this algorithm needs the original input signal, which is not flexible

From the extraction scheme, frequency masking is a private watermarking.

Page 14: Audio watermarking

SPREAD SPECTRUM

Wherewi is element of watermark y is the watermarked signal

A correlation detector is carried out to detect the presence of the watermark and is described by Eqn

where N is vector cardinalityCont..

Page 15: Audio watermarking

SPREAD SPECTRUM

While x can be modelled as Gaussian random vector, the normalised correlation test is described as Eqn

10, /

^ 2 x

CQ p N N

N

If p = 1, watermark is presentIf p=0, watermark is not present

Page 16: Audio watermarking

ANALYSIS OF VARIOUS TECHNIQUES DC level shifting algorithm

The robustness level for the DC level shifting is poor, the watermark

can be easily removed by altering the DC levels of the signal and the watermark is

destroyed.

Frequency Masking

This method is not robust to filtering because watermark is embedded

to all frequency bands.

Attack by high pass filter MPEG can erase the watermark.

Spread Spectrum & Band Division Based On QMF

Both the band division method and the spread spectrum method are

capable of defending against the MP3 compression, but the spread spectrum

has a better detection rate than the band division method.

Cont..

Page 17: Audio watermarking

The implementation steps for spread spectrum are more complicated. This

makes it hard to hack and breakdown the algorithm.

Band Division Based On QMF

Band division technology performs extraction scheme with no original

signal require attacks are still not upgraded to hack down this audio watermarking

technology.

Page 18: Audio watermarking

CONCLUSION Watermark can be embedded in an audio signal without

being recognized by hearing and the perceptibility requirement is fulfilled .

Extraction can be done without the presence of the original signal and the watermark itself, so can be classified as blind or public watermarking.

Thank You !!!

Page 19: Audio watermarking

QUESTIONS??