System and method of retrieving a watermark within a signal

• US 7,131,007 B1
• Filed: 03/26/2002
• Issued: 10/31/2006
• Est. Priority Date: 06/04/2001
• Status: Expired due to Fees

• Current Assignee: AT&T Corporation (AT&T, Inc.)
• Original Assignee: AT&T Corporation (AT&T, Inc.)
• Inventors: Johnston, James David, Kuo, Shyh-Shiaw, Turin, William, Quackenbush, Schuyler Reynier
• Primary Examiner(s): Sheikh, Ayaz
• Assistant Examiner(s): Shiferaw, Eleni
• Application Number:
  US10/107,017
• Time in Patent Office: 1,680 Days
• Field of Search: 380/282, 380/252, 713/176
• US Class Current: 713/173
• CPC Class Codes: G10L 19/018 Audio watermarking, i.e. embedding inaudible data in the audio signal

Abstract

A system and method of retrieving a watermark in a watermarked signal are disclosed. The watermarked signal comprises odd and even overlapped blocks where the watermark is contained in the even blocks. The method comprises, for each k-th even block, subtracting the two adjacent odd blocks from the k-th even block of the watermarked signal to retrieve {overscore (s)}*_k(n), transforming {overscore (s)}*_k(n) into the frequency domain to generate {overscore (S)}_k(f), calculating a phase of {overscore (S)}_k(f) as {overscore (φ)}(f) and a phase of S_k(f) as φ(f), calculating the difference Ψ(f) between {overscore (φ)}(f) and φ(f), unwrapping Ψ(f) to obtain the phase modulation {tilde over (Φ)}_k(f), and using a Viterbi search to retrieve the watermark embedded in {tilde over (Φ)}_k(f).

Claims

1. A computer-implemented method of retrieving a watermark in a watermarked signal, the watermarked signal comprising odd and even overlapped blocks where the watermark is contained in even blocks, the method comprising, for each k-th block:

subtracting odd blocks from a k-th block of the watermarked signal to generate {overscore (s)}*_k(n);

applying an FFT to {overscore (s)}*_k(n) to generate a phase {overscore (S)}_k(f);

calculating a phase of {overscore (S)}_k(f) as {overscore (φ

)}(f) and a phase of an original signal S_k(f) as φ

(f);

calculating the difference Ψ

(f) between {overscore (φ

)}(f) and φ

(f); and

using a Viterbi search to retrieve the watermark embedded in Ψ

(f), wherein if during a phase-modulation stage of generating the watermarked signal, the result of adding a phase-modulation to the phase of the original signal has an absolute value greater than π

, then the method further comprises;

unwrapping Ψ

(f) to obtain a correct phase modulation {tilde over (Φ

)}_k(f) only when φ

(f)>

π

/2 and Ψ

(f) is greater than a dynamic range of the phase modulation; and

using the Viterbi search to retrieve the watermark embedded in {tilde over (Φ

)}_k(f).

2. A method of retrieving a watermark in a watermarked signal of claim 1, wherein odd blocks subtracted from the k-th even block are the two adjacent odd blocks of the original signal to the k-th even block.

3. A method of retrieving a watermark in a watermarked signal of claim 1, wherein the watermarked signal is an audio signal.

4. A computer-implemented method of retrieving a watermark embedded in a watermarked signal, the watermarked signal comprising odd and even overlapped blocks where the watermark is contained in even blocks and wherein the absolute value of adding a phase modulation Φ

_k(f) to a phase of an original signal in a phase-modulation step of generating the watermarked signal is greater than π

, the method comprising, for each k-th block of the watermarked signal;

subtracting odd blocks from a k-th block to generate {overscore (s)}*_k(n);

applying an FFT to {overscore (s)}*_k(n) to generate a phase {overscore (S)}_k(f);

calculating a phase of {overscore (S)}*_k(f) as {overscore (φ

)}(f) and a phase of an original signal S_k(f) as φ

(f)

calculating the difference Ψ

(f) between {overscore (φ

)}(f) and φ

(f);

unwrapping Ψ

(f) to generate {tilde over (Φ

)}_k(f), which contains the embedded watermark,

wherein the unwrapping only occurs when φ

(f)>

π

/2 and Ψ

(f) is greater than a dynamic range of a phase modulation.

5. The method of retrieving a watermark embedded in a watermarked signal of claim 4, further comprising:

using a Viterbi search to retrieve the watermark embedded in {tilde over (Φ

)}_k(f).

6. A computer-implemented method of retrieving a watermark embedded in a watermarked signal, the method using the phase S_k(f) of an original signal, the watermarked signal comprising odd and even overlapped blocks where the watermark is contained in even blocks, the method comprising, for each k-th even block:

(a) subtracting two adjacent odd blocks from a k-th even block of the watermarked signal to retrieve {overscore (s)}*_k(n);

(b) transforming {overscore (s)}*_k(n) into a frequency domain to generate {overscore (S)}_k(f);

(c) calculating a phase of {overscore (S)}_k(f) as {overscore (φ

)}(f) and a phase of S_k(f) as φ

(f);

(d) calculating the difference Ψ

(f) between {overscore (φ

)}(f) and φ

(f);

(e) unwrapping Ψ

(f) to obtain the phase modulation {tilde over (Φ

)}_k(f) only if, during the phase-modulation step of generating the watermarked signal, the absolute value of the result of adding a phase modulation Φ

_k(f) to a phase of the original signal is greater than π

, when φ

(f)>

π

/2 and when Ψ

(f) is greater than the dynamic range of the phase modulation; and

(f) using a Viterbi search to retrieve the watermark embedded in {tilde over (Φ

)}_k(f).

7. A method of retrieving a watermark in a watermarked signal generated from an original signal of claim 6, wherein the watermarked signal is an audio signal.

8. A computer-implemented method of retrieving a watermark embedded in a watermarked signal, the method using the phase S_k(f) of an original signal, the watermarked signal comprising odd and even overlapped blocks where the watermark is contained in even blocks, the method comprising, for each k-th even block:

obtaining a phase modulation {tilde over (Φ

)} _k(f) within a k-th even block; and

performing a Viterbi search using an energy-weighted mean absolute error L₁ norm to retrieve the watermark embedded in {tilde over (Φ

)} _k(f), wherein the method further comprises using the following cost function associated with the L₁ norm when performing the Viterbi search;

$<br><br>c_{\mathrm{ij}}<br><br>\left(t\right)=<br><br>\frac{1}{K}<br><br>\underset{f=0}{\overset{K-1}{\sum <br><br>}}<br><br><br><br><br><br>\underset{c}{\sum <br><br>}<br><br><br><br>\left(p_{\mathrm{ij}}<br><br>\left(f\right)-o_t<br><br>\left(f\right)\right)<br><br>w_t<br><br>\left(f\right)<br><br>,<br><br>\mathrm{for}<br><br><br><br>\left(\begin{array}{cc}0\le <br><br>i,& j\le <br><br>1\\ 1\le <br><br>t\le <br><br>T,& \end{array}\right),$

where f_ij(f) is the path template between state i and j, K is the total number of frequency bins associated with the observation o_t, and w_t(f) are the weights which are based on spectrum energy.

9. The method of retrieving a watermark embedded in a watermarked signal of claim 8, wherein w_f(f) are the weights that are defined as:

$w_t<br><br>\left(f\right)=\min <br><br>\left({<br><br>S^{\mathrm{\prime <br><br>}}<br><br>\left(f\right)<br><br>}^2,{<br><br>{{\stackrel{\_}{S}}_c}^{\mathrm{\prime <br><br>}}<br><br>\left(f\right)<br><br>}^2\right),\mathrm{for}<br><br><br><br>f=0,\mathrm{\dots <br><br>}<br><br>,K-1$ $\underset{f}{\sum <br><br>}<br><br><br><br>w_t<br><br>\left(f\right)=1.$

10. The method of retrieving a watermark embedded in a watermarked signal of claim 8, wherein the signal is a multi-channel signal.

11. The method of retrieving a watermark in a watermarked signal of claim 10, further comprising:

using the following cost function and spectrum energy weights associated with the L₁ norm when performing the Viterbi search;

$c_{\mathrm{ij}}<br><br>\left(t\right)=<br><br>\frac{1}{K}<br><br>\underset{f=0}{\overset{K-1}{\sum <br><br>}}<br><br><br><br><br><br>\underset{c}{\sum <br><br>}<br><br><br><br>\left(p_{\mathrm{ij}}<br><br>\left(f\right)-o_{t,c}<br><br>\left(f\right)\right)<br><br>w_{t,c}<br><br>\left(f\right)<br><br>,<br><br>\mathrm{for}<br><br><br><br>\left(\begin{array}{cc}0\le <br><br>i,& j\le <br><br>1\\ 1\le <br><br>t\le <br><br>T,& \end{array}\right),<br><br>w_{\mathrm{tc}}<br><br>\left(f\right)=\min <br><br>\left({<br><br>{S_c}^{\mathrm{\prime <br><br>}}<br><br>\left(f\right)<br><br>}^2,{<br><br>{{\stackrel{\_}{S}}_c}^{\mathrm{\prime <br><br>}}<br><br>\left(f\right)<br><br>}^2\right),<br><br>\mathrm{for}<br><br><br><br>\left(\begin{array}{cc}f=0,\mathrm{\dots <br><br>}& K-1\\ c=1,\mathrm{\dots <br><br>}& M<br><br>\left(\mathrm{Totalchannels}\right)\end{array}\right),<br><br>\underset{f}{\sum <br><br>}<br><br>\underset{c}{\sum <br><br>}<br><br>w_{t,c}<br><br>\left(f\right)=1.$

12. The method of retrieving a watermark in a watermarked signal of claim 10, further comprising:

(a) initializing parameters C₁(i)=c_ii, i=0, 1 and γ

_t(i)=0;

(b) using recursion to calculate;

$C_t<br><br>\left(j\right)=\underset{i=1,2}{\min }<br><br>\left[C_{t-1}<br><br>\left(i\right)+c_{\mathrm{ij}}<br><br>\left(t\right)\right],2\le <br><br>t\le <br><br>T,j=-0,2\le <br><br>t\le <br><br>T,j=-0,1$ $\mathrm{\gamma <br><br>}<br><br><br><br>t<br><br>\left(j\right)=\arg <br><br><br><br>\underset{i=1,2}{\min }<br><br>\left[C_{t-1}<br><br>\left(i\right)+c_{\mathrm{ij}}<br><br>\left(t\right)\right],2\le <br><br>t\le <br><br>T,j=-0,2\le <br><br>t\le <br><br>T$

(c) using the following calculations to determine the minimum total cost associated with a best state sequence q;

$\begin{array}{c}{C}^{*}=\underset{i=0,1}{\min }<br><br>\left[C_T<br><br>\left(i\right)\right]\\ q_T=\arg <br><br><br><br>\underset{i=0,1}{\min }<br><br>\left[C_T<br><br>\left(i\right)\right]\end{array};<br><br>\mathrm{and}$

(d) state sequence backtracking to calculate;

q_t=γ

_t+1(q_t+1),t=T−

1,T−

2, . . . ,1.

13. The method of retrieving a watermark in a watermarked signal of claim 8, further comprising:

(a) initializing parameters C₁(i)=c_ii, i=0, 1 and γ

_t(i)=0;

(b) using recursion to calculate;

$C_t<br><br>\left(j\right)=\underset{i-1,2}{\min }<br><br>\left[C_{t-1}<br><br>\left(i\right)+c_{\mathrm{ij}}<br><br>\left(t\right)\right],2\le <br><br>t\le <br><br>T,j=-0,2\le <br><br>t\le <br><br>T,j=-0,1$ ${\mathrm{\gamma <br><br>}}_t<br><br>\left(j\right)=\arg <br><br><br><br>\underset{i=1,2}{\min }<br><br>\left[C_{t-1}<br><br>\left(i\right)+c_{\mathrm{ij}}<br><br>\left(t\right)\right],2\le <br><br>t\le <br><br>T,j=-0,2\le <br><br>t\le <br><br>T$

(c) using the following calculations to determine the minimum total cost associated with a best state sequence q;

$\begin{array}{c}{C}^{*}=\underset{i=0,1}{\min }<br><br>\left[C_T<br><br>\left(i\right)\right]\\ q_T=\arg <br><br><br><br>\underset{i=0,1}{\min }<br><br>\left[C_T<br><br>\left(i\right)\right]\end{array};<br><br>\mathrm{and}$

(d) using the following to calculate state sequence backtracking;

q_t=γ

_t+1(q_t+1),t=T−

1,T−

2, . . . ,1.