Combined Watermarking and Fingerprinting

• US 20100322469A1
• Filed: 05/21/2010
• Published: 12/23/2010
• Est. Priority Date: 05/21/2009
• Status: Active Grant

• Current Assignee: Digimarc Corporation
• Original Assignee: Digimarc Corporation
• Inventors: Sharma, Ravi K.
• Granted Patent: US 8,300,884 B2
• Time in Patent Office: Days
• US Class Current: 382/100
• CPC Class Codes: G06T 1/0028 Adaptive watermarking, e.g. Human Visual System [HVS]-based watermarking
• CPC Class Codes: G10L 19/018 Audio watermarking, i.e. embedding inaudible data in the audio signal

Abstract

Content fingerprints and watermarks are combined in various ways for content identification applications. Fingerprints are used to identify content generally while watermarks provide more detailed localization of parts within the content, and vice versa. Fingerprint techniques are further used for signal synchronization and other pre-processing steps to assist in digital watermark decoding. A variety of fingerprint/watermark techniques identify characteristics of the channel of content from content samples.

Claims

1. A method of identifying a content signal comprising:

using a programmed processor to perform the acts of;

decoding a digital watermark from a received portion of the content signal, including aggregating instances of the digital watermark repeated in the content signal;

the digital watermark serving to identify the content signal and associated content fingerprints for the content signal;

using the content fingerprints to localize the received portion of the content signal.

2. The method of claim 1 wherein using the content fingerprints to localize the received portion comprises:

extracting fingerprints from the received portion of the content signal; and

matching extracted fingerprints from the received portion of the content signal with the associated content fingerprints to determine which part of the content signal the digital watermark was detected from.

3. A method of identifying a content signal comprising:

using a programmed processor to perform the acts of;

computing a fingerprint for a received portion of the content signal;

using the fingerprint to identify the content signal; and

extracting a digital watermark from the received portion of the content signal to determine localization information for content within the received portion.

4. The method of claim 3 wherein the content signal includes plural unique digital watermarks corresponding to locations within the content signal, and the localization information is determined from unique information in the extracted digital watermark corresponding to a location within the content signal.

5. The method of claim 4 wherein the unique information comprises a time code providing a temporal location in the content signal.

6. A content recognition method comprising:

using a programmed processor to perform the acts of;

computing a fingerprint for a received portion of a content signal;

sending the fingerprint to a reference database to identify the content signal, the reference database providing identity by matching the fingerprint with reference fingerprints in the database;

based on the identity, obtaining a reference copy of the content signal; and

determining distortion of the received portion of the content signal based on a comparative analysis between the received portion and the reference copy.

7. The content recognition method of claim 6 wherein the reference copy and the received portion are used to determine the geometric transformation of the received portion relative to the reference copy.

8. The content recognition method of claim 7 including applying the geometric transformation to synchronize the received portion for extraction of a digital watermark from the received portion.

9. The content recognition method of claim 8 including:

extracting the digital watermark from the received portion.

10. The content recognition method of claim 6 wherein the distortion is used to determine a distribution channel of the received portion.

11. A method of content recognition comprising:

using a programmed processor to perform the acts of;

analyzing a received portion of a content signal to infer signal degradation of the content signal;

determining a distribution channel from the signal degradation;

computing a fingerprint for the received portion of the content signal;

sending the fingerprint to a reference database to identify the content signal, the reference database providing identity by matching the fingerprint with a set of reference fingerprints in the database corresponding to the distribution channel, the database being segmented into different sets of reference fingerprints corresponding to distribution channels such that the different sets include reference fingerprints for a common set of content signals, but store different reference fingerprints for the common set of content signals that are computed based on signal characteristics relevant for the different distribution channels; and

using the distribution channel to determine which of the different sets to use to identify the content signal.

12. The method of claim 11 wherein the distribution channel corresponds to compression format.

13. A content recognition method comprising:

using a programmed processor to perform the acts of;

receiving different versions of a content signal, the different versions corresponding to different distribution channels;

computing reference fingerprints for the different versions, and storing the reference fingerprints in a corresponding section of a segmented database, the corresponding section being associated with a particular distribution channel; and

in response to a request, determining a match for a received fingerprint by searching for a match in a section of the database corresponding to distribution channel associated with the received fingerprint.

14. A method of digital watermark decoding comprising:

receiving a watermarked electronic signal in which a watermark having first and second components, where the first component is repeated over a first temporal portion of the electronic signal, and the second component is not repeated over the first temporal portion;

selectively accumulating the first component; and

decoding the watermark from the accumulated first component and the second component from the first temporal portion.

15. The method of claim 14 wherein the decoding comprises performing error correction decoding on the accumulated first component and the second component.

16. The method of claim 15 wherein the error correction decoding comprises convolutional decoding.

17. The method of claim 14 including decoding the watermark using a weighted repetition scheme wherein a component of the watermark that is repeated more frequently is weighted more than a component repeated less frequently in the electronic signal.

18. The method of claim 14 including selectively combining coded message symbols that are common between adjacent watermark segments in the watermarked electronic signal.

19. The method of claim 14 including:

accumulating a common component and a varying component of a watermark signal between adjacent watermark segments in the watermarked electronic signal; and

interpolating between common components to decode the varying component.

20. The method of claim 17 including:

decoding an implicit weighted repetition code by accumulating common coded message symbols of adjacent segments; and

decoding an explicit weighted repetition code in which a portion of a message payload of the digital watermark is given higher repetition weighting in the error correction coding of the message payload.