Sei sulla pagina 1di 32

AUDIO WATERMARKING

Submitted By: CHAITHANYA BS PRIYANKA L HARSHA S ANJU SURESH

INTRODUCTION
With the growth of multimedia production and

broadcasting, security issues have been concerning for multimedia producers.


Watermarking has provided good means for

protecting multimedia against such illegal tasks.


But watermarking works have been dedicated to image

watermarking so far.

Watermarking
is a method of protecting multimedia data against

unauthorized copying It is a signal added to the original signal, which can later be extracted or detected. It is intended to be permanently embedded into the digital data watermark should not degrade the quality of the digital data.

Watermarking
Time domain
watermark bits are embedded directly

Frequency Domain
Embedded into the transform coefficients

REQUIREMENTS
Imperceptibility Payload

Security
Robustness

Watermarking algorithms aim at achieving suitable

trade-offs among these requirements.

EMBEDDING(DWT)

Audio

DWT

Embedding Watermark

IDWT

Watermarked Audio

DISCRETE WAVELET TRANSFORM


Discrete wavelet transform (DWT) decompose signals

into sub bands with smaller bandwidths and lower sample rates
basic computational step is a convolution followed by a

decimation.

Multilevel wavelet decomposition


X
cA1 cD1

cA2

cD2

cA3

cD3 -cAi-Approximation Coeff -cDi-Detailed Coeff

Quantization Index Modulation


One of the common techniques used for watermarking Achieve provably good rate-distortion-robustness

performance to embed a 1 bit, the quantized value should be odd and to embed a 0 bit, the value should be even.

EMBEDDING ALGORITHM(DWT)
START Frames of 1024 samples are taken

Decompose into 5 levels using DWT

de1=d1d1T de2=d2d2T

ep=de1/de2

ep=sf*ep

Round ep to nearest odd no ep

NO

W=0

YES

Round ep to nearest even no ep

K=(ep/ep)

d2=kd2

Water marked Audio signal

EXTRACTION

Water marked signal

DWT

Extraction

Extracted Watermark

Extraction Algorithm(DWT)
START Frames of 1024 samples are taken

Decompose into 5 levels using DWT

dew1=dw1dw1T dew2=dw2dw2T

epw=dew1/dew2

epw=sf*epw

epw=round(epw)

Extracted bit is 0

NO

ep=even

YES

Extracted bit is 1

STOP

SVD
SVD is an optimal matrix-decomposition technique

It packs maximum signal energy into a few coefficients

as possible
has the ability of adapting to variations have high payload

Dither-modulation(DM)
has good performance in terms of robustness,

imperceptibility, data payload and blind extraction


the host signal is dithered using watermark

information

EMBEDDING(SVD)

Audio

SVD

Embedding Watermark

Inverse SVD

Watermarked Audio

EMBEDDING ALGORITHM(SVD)
Original Audio X Frames of 1024 samples are taken

Segmenting into Blocks Bj

2D Matrix Formation

Apply SVD

Compute n and integer a

a = a + d ((a + d) mod (2d))

NO

w(i,j)=1

YES

a = a + d (a mod (2d)

Compute n and modified SVs Apply Inverse SVD

Segment Reconstruction

EXTRACTION(SVD)

Water marked signal

SVD

Extraction

Extracted Watermark

Extraction Algorithm
Watermarked Audio Frames of 1024 samples are taken

Segmenting into Blocks B j

2D Matrix Formation

Apply SVD

Compute n and integer a

NO

YES
mod(a,2)=0

w(i, j) = 0

w(i, j) = 1

Watermark Bits Formation

Contd
The method is simple, has low complexity, and

achieves good rate-distortion-robustness trade-offs


Changing SVs slightly does not affect the signal quality
SVs do not change much after various types of

common signal processing operations

PROPOSED METHOD OF WATERMARKING


Uses both DWT and SVD for embedding and

extracting watermark
The second Detailed Coefficient of DWT is segmented

into blocks and SVD is taken on this cD2

EMBEDDING

Audio

DWT

SVD

Embedding Watermark

Inverse SVD

Watermarked Audio

Flowchart

Original Audio X

Frames of 1024 samples are taken

Take DWT

Segmenting cD2 into Blocks B j

2D Matrix Formation

Apply SVD

Compute n and integer a

a = a + d ((a + d) mod (2d))

NO

w(i,j)=1

YES

a = a + d (a mod (2d)

Compute n and modified SVs Apply Inverse SVD

Segment Reconstruction

EXTRACTION

Water marked signal

DWT

SVD

Extraction

Extracted Watermark

Flowchart
Watermarked Audio Frames of 1024 samples are taken

Take DWT

Segmenting cD2 into Blocks B j

2D Matrix Formation

Apply SVD

Compute n and integer a

NO

YES
mod(a,2)=0

w(i, j) = 0

w(i, j) = 1

Watermark Bits Formation

CONCLUSION
By the proposed method we achieve infinite peak

signal to noise ratio for the watermarked signal.


Obtained PSNR for the embedded signal is 89.7997 Thus obtained an optimal Watermarking Techniques.

Thank You!!

Potrebbero piacerti anche