Method and System for Blind Audio Watermarking

• US 20140005815A1
• Filed: 09/10/2012
• Published: 01/02/2014
• Est. Priority Date: 06/27/2012
• Status: Active Grant

• Current Assignee: TATA Consultancy Services Limited (Tata Sons Pvt Ltd.)
• Original Assignee: Krishna Rao Kakkirala, Srinivasa Rao Chalamala
• Inventors: Chalamala, Srinivasa Rao, Kakkirala, Krishna Rao
• Granted Patent: US 9,269,362 B2
• Time in Patent Office: Days
• US Class Current: 700/94
• CPC Class Codes: G06F 17/148 Wavelet transforms
• CPC Class Codes: G10L 19/018 Audio watermarking, i.e. embedding inaudible data in the audio signal

Abstract

A method and system for blind audio watermarking has been envisaged. The system embeds an in-audible watermark by modifying a predetermined singular value in the wavelet domain. The generated watermarked audio signals are robust again Gaussian noise and other watermark removal and compression attacks. Also, the system can perform watermark extraction/detection in real-time without requiring in advance, either the original audio signal or its processed parameters. Thereby, the method and system provide a real-time, efficient technique for blind audio watermarking.

Claims

1. A method for performing at least one of embedding and extraction of audio watermark, the method comprising the following steps:

segmenting an input audio signal into segments of fixed size;

decomposing each of said segments using multilevel Discrete Wavelet Transform (DWT) to retrieve detailed coefficients;

forming a ‘

detail coefficient matrix’

of said detailed coefficients retrieved for each level of said decomposition; and

applying Singular Value Decomposition (SVD) on said ‘

detail coefficient matrix’

to obtain three decomposed matrices including a left singular vector matrix, a right singular vector matrix and a singular value matrix containing four singular values S1, S2, S3 and S4.

2. The method as claimed in claim 1, wherein the method of embedding the watermark includes the following steps:

embedding a watermark bit in said segment of said input audio signal by modifying said S2 value of said singular value matrix using a combination of said S1 value, said S4 value and at least two scaling factors having pre-designated value alpha1 and alpha2 based on the formula S2=alpha1*(S1+S4)+alpha2*watermark bit*(S1−

S4);

obtaining a modified ‘

detail coefficient matrix’

by performing inverse SVD operation by combining the modified singular value matrix with said unmodified left singular vector matrix, and said unmodified right singular vector matrix;

performing multilevel Inverse Discrete Wavelet Transform (IDWT) on said modified ‘

detail coefficient matrix’

;

iterating aforementioned steps until discrete watermark bits are embedded in each of said segments; and

obtaining a watermarked audio signal by combining each of the modified segments.

3. The method as claimed in claim 1, wherein the method of extraction the watermark includes the following steps:

determining a threshold value using said S1 value, said S4 value and a scaling factor having pre-designated value alpha1 using the formula Threshold value=(alpha1*(S1+S4));

extracting a watermark bit from each of said segments by comparing the value of S2 with said threshold value using the formula;

watermark bit=0, if S2<

Threshold Value and

watermark bit=1, if S2>

=Threshold Value;

scrambling and error correction to eliminate errors in extraction of said watermark bit; and

combining watermarked bits extracted from each of said segments to form the watermark.

4. The method as claimed in claim 1, wherein the step of decomposing each of said segments using multilevel Discrete Wavelet includes the step of transforming said segments of said input audio signal in the Discrete Wavelet Transform (DWT) domain using a four-level discrete wavelet decomposition technique selected from the group consisting of Haar, Daubechies, Coiflet, Legendre and symlets.

13. The system as claimed in claim 1, wherein said watermark extraction unit includes error correction means to use scrambling and error correction codes to eliminate errors in watermark extraction to improve accuracy.

5. A computer-implemented system for audio watermarking comprising:

a watermark embedding unit to embed an in-audible watermark in an input audio signal by modifying a predetermined singular value derived by performing Singular Value Decomposition (SVD) operation on said input audio signal in the Discrete Wavelet Transform (DWT) domain; and

a watermark extraction unit to perform said SVD operation on a received watermarked audio signal in the DWT domain to derive singular values, said watermark extraction unit comprising an analyzer to identify an embedded watermark by comparing the derived singular values,

wherein four singular values S1, S2, S3 and S4 are derived on performing said SVD operation on a received input audio/watermarked audio signal in the DWT domain.

6. The system as claimed in claim 5, wherein said watermark is represented in the bipolar format and selected from the group consisting of a logo image, a barcode and a random binary sequence.

7. The system as claimed in claim 5, wherein said watermark embedding unit and said watermark extraction unit comprise discrete processing means including Media Processors and Digital Signal Processors to perform signal processing and matrix-based operations for watermark embedding and extraction.

8. The system as claimed in claim 5, wherein said watermark embedding unit and said watermark extraction unit is adapted to perform a four-level discrete wavelet decomposition using a technique selected from the group consisting of Haar, Daubechies, Coiflet, Legendre and symlets.

9. The system as claimed in claim 5, wherein said watermark embedding unit and said watermark extraction unit derive said four singular values S1, S2, S3 and S4 by performing at least a four-level DWT operation on predetermined fixed-size segments of said input audio signal and further retrieving three decomposed matrices including a left singular vector matrix, a right singular vector matrix and a singular value matrix containing four singular values S1, S2, S3 and S4, wherein said coefficients of a DWT level are arranged to form a ‘

detail coefficient matrix’

, wherein each level forms a row in the matrix for a particular segment.

10. The system as claimed in claim 5, wherein said watermark embedding unit modifies said S2 value by replacing the value of said S2 by a value obtained by combining said S1 value, said S2 value, a watermark bit and at least two scaling factors having pre-designated values alpha1 and alpha2 based on the formula S2=alpha1*(S1+S4)+alpha2*watermark bit*(S1−

S4).

11. The system as claimed in claim 5, wherein said watermark embedding unit obtains a watermarked audio signal by applying inverse SVD and inverse DWT techniques on each of said segments of said input audio signal and combining the resultant segments.

12. The system as claimed in claim 5, wherein said watermark extraction unit is adapted to extract said watermark bits in segments of said watermarked audio signal by computing a threshold value based on the formula (alpha1*(S1+S4)) and said analyzer compares S2 value with said threshold value.