METHOD AND APPARATUS FOR EMBEDDING WATERMARKS

• US 20090074240A1
• Filed: 11/12/2008
• Published: 03/19/2009
• Est. Priority Date: 06/13/2003
• Status: Active Grant

• Current Assignee: Nielsen Company (US), LLC
• Original Assignee: Nielsen Company (US), LLC
• Inventors: Srinivasan, Venugopal
• Granted Patent: US 7,643,652 B2
• Time in Patent Office: Days
• US Class Current: 382/100
• CPC Class Codes: G06T 1/0021 Image watermarking
• CPC Class Codes: G10L 19/018 Audio watermarking, i.e. embedding inaudible data in the audio signal
• CPC Class Codes: G10L 19/0212 using orthogonal transformation
• CPC Class Codes: G10L 25/27 characterised by the analysis technique

Abstract

Methods and apparatus for embedding a watermark are disclosed. An example method to embed media identification information in a compressed media data stream disclosed herein comprises reconstructing an uncompressed media data stream from the compressed media data stream, the uncompressed media data stream separate from the compressed media data stream, embedding the media identification information in the uncompressed media data stream to determine a watermarked uncompressed media data stream, and modifying a first mantissa value corresponding to a first transform coefficient associated with the compressed media stream to embed the media identification information in the compressed media data stream without uncompressing the compressed media data stream, the modification of the first mantissa value based on a difference between the first transform coefficient and a corresponding second transform coefficient generated from the watermarked uncompressed media data stream.

Claims

1. A method to embed media identification information in a compressed media data stream, the method comprising:

reconstructing an uncompressed media data stream from the compressed media data stream, the uncompressed media data stream separate from the compressed media data stream;

embedding the media identification information in the uncompressed media data stream to determine a watermarked uncompressed media data stream; and

modifying a first mantissa value corresponding to a first transform coefficient associated with the compressed media stream to embed the media identification information in the compressed media data stream without uncompressing the compressed media data stream, the modification of the first mantissa value based on a difference between the first transform coefficient and a corresponding second transform coefficient generated from the watermarked uncompressed media data stream.

2. A method as defined in claim 1 wherein the compressed media data stream comprises a compressed audio data stream and the uncompressed media data stream comprises a time-domain audio data stream.

3. A method as defined in claim 1 wherein the media identification information comprises a watermark representing at least one of program or source identification information.

4. A method as defined in claim 1 wherein the first and second transform coefficients comprise respective first and second modified discrete cosine transform (MDCT) coefficients.

5. A method as defined in claim 1 wherein reconstructing the uncompressed media data stream from the compressed media data stream comprises:

performing an inverse transform on a copy of the compressed media data stream to determine first and second inverse transformed data blocks; and

combining the first and second inverse transformed data blocks to form the uncompressed media data stream.

6. A method as defined in claim 1 wherein embedding the media identification information in the uncompressed media data stream to determine the watermarked uncompressed media data stream comprises:

increasing a first frequency component of the uncompressed media data stream and decreasing a second frequency component of the uncompressed media data stream to represent a first data value associated with the media identification information; and

decreasing the first frequency component of the uncompressed media data stream and increasing the second frequency component of the uncompressed media data stream to represent a second data value associated with the media identification information.

7. A method as defined in claim 1 wherein embedding the media identification information in the uncompressed media data stream to determine the watermarked uncompressed media data stream comprises:

determining a plurality of code signal components to represent a data value associated with the media identification information; and

combining the plurality of code signal components and the uncompressed media data stream based on a determined masking energy.

8. A method as defined in claim 1 wherein modifying the first mantissa value corresponding to the first transform coefficient associated with the compressed media stream comprises:

determining a second mantissa value associated with the second transform coefficient generated from the watermarked uncompressed media data stream;

quantizing the second mantissa value based on compression information associated with the first mantissa value; and

replacing the first mantissa value with the quantized second mantissa value.

9. A method as defined in claim 1 wherein the first transform coefficient further comprises the first mantissa value and a first exponent value, and wherein the first exponent value is not modified to embed the media identification information.

10. A method as defined in claim 9 wherein the first mantissa value is set to at least one of a minimum or maximum value based on compression information associated with the first mantissa value when modification of only the first mantissa value is not sufficient to account for the difference between the first transform coefficient and the second transform coefficient.

11. A method of embedding a watermark comprising:

identifying a transform coefficient associated with a compressed digital data stream;

obtaining a mantissa code; and

modifying the transform coefficient based on the mantissa code to embed the watermark.

12. A method as defined in claim 11 wherein the mantissa code is a first mantissa code, and wherein modifying the transform coefficient comprises substituting the first mantissa code for a second mantissa code associated with the transform coefficient.

13. A method as defined in claim 11 wherein obtaining the mantissa code comprises:

selecting a code signal frequency to encode to the transform coefficient based on data to be embedded;

obtaining a masking energy associated with the code signal frequency to be encoded to the transform coefficient;

selecting a magnitude for a watermarked transform coefficient associated with the mantissa code based on the masking energy; and

determining the mantissa code based on the magnitude.

14. A method as defined in claim 13 wherein the code signal frequency comprises one or more sinusoidal components, and wherein each sinusoidal component includes a frequency based on a desired code.

15. A method as defined in claim 11 wherein modifying the transform coefficient comprises:

generating a time-domain audio block based on the transform coefficient;

generating a reconstructed audio block based on the time-domain audio block; and

generating a watermarked audio block based on the reconstructed audio block.

16. A method as defined in claims 11 wherein modifying the transform coefficient comprises generating a watermarked transform coefficient based on a first watermarked audio block and a second watermarked audio block.

17. A method as defined in claim 16 wherein generating the watermarked coefficient based on the first and second watermarked audio blocks comprises obtaining the mantissa code based on compression information associated with the compressed digital data stream.

18. A method to determine ratings information comprising:

extracting identification information embedded in presented media content, the identification information embedded in a broadcasted compressed audio data stream corresponding to the presented media content, the identification information embedded in the compressed audio data stream without uncompressing the compressed audio data stream by;

modifying a first mantissa value corresponding to a first transform coefficient associated with the compressed audio data stream to embed the identification information into the compressed audio data stream, the first mantissa value to be modified based on a difference between the first transform coefficient and a corresponding second transform coefficient generated from a separate uncompressed version of the compressed audio data stream into which the identification information has also been embedded.

19. A method as defined in claim 18 wherein the identification information comprises a watermark representing at least one of program or source identification information.

20. A method as defined in claim 18 further comprising:

uncompressing the broadcasted compressed audio data stream at a receiving device to generate an uncompressed audio data stream corresponding to the presented media content; and

extracting the identification information from an analog audio signal corresponding to the uncompressed audio data stream, the analog audio signal provided by at least one of an analog output or a speaker of the receiving device.