System for low-latency detection of known audio video content using audio fingerprinting and audio watermarking

1. A system for low-latency detection of known audio-video content in multiple audio-video streams, received over a transport multiplex wherein multiple advertisements of different durations are being transmitted over a network, having (a) an ingest module, (b) at least one repository of content assets, (c) at least one repository of advertisement assets, (d) a playout server, (e) at least one repository of advertisement assets, and (f) at least one network streams, comprising:

a) a Watermarking and Fingerprinting (WMFP) module, which receives multiple advertisements A having audio-video streams, from a Master AV Database, placing an audio watermark on the incoming streams and extracting a high-confidence fingerprint (signature) from each of the watermarked assets A′;

b) communicating fingerprint signatures and replacement rules to at least one receiver over the Internet;

c) at least one repository of local content assets; and

d) at least one receiver at different locations, actively replacing detected advertisements having AV content, using a signature detector, with local content assets, based on the replacement rules,

wherein the at least one receiver has low complexity signature detectors further comprising:

a) the detector receiving a broadcast AV stream for which it keeps generating the fingerprint for a window of audio, which has a fixed match duration;

b) matching the generated signature against all the references stored in the signature DB;

c) representing the signatures as a sequence of bit, a bit string;

d) detecting a match when the generated signature and a reference signature have the number of bits matching above a predefined threshold T; and

e) disambiguating at least two similar sounding master AVs having a match more than T by considering the master AV stream having the maximum matched bits compared to the current audio snapshot signature as the right master AV stream match.

2. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the WMFP module places an audio watermark on the Master AV stream, the audio watermark being a pseudo-random, inaudible noise sequence.

3. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the advertisement assets from the Master AV Database has similar audio portions over a certain duration of their content.

4. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the WMFP module extracts a fingerprint signature from the watermarked content, the fingerprint being unique across assets with similar audio patterns toward the beginning due to the pseudo-random watermark, thereby disambiguating assets with similar sounding audio over an initial match duration at the receiver.

5. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the replacement rules specify an identifier for the Master AV stream, the corresponding fingerprint signature and the replacement schedule on specific local receivers.

6. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the at least one repository of local content assets comprise content assets that are specific to certain demographics, geographies, and customers.

7. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the at least one receiver has signature detectors further comprising:

a) performing watermark verification by extracting residual noise audio from incoming audio;

b) correlating the extracted noise against the pseudo-random, inaudible noise sequence generated by the WMFP module; and

c) detecting a match when the Master AV and the watermarked output match.

8. The system for low-latency detection of known audio-video content in multiple audio-video streams of claim 1 , wherein the receivers at different locations monitor the transmitted content, actively replacing detected advertisements (A′) using the signature detector, with local content assets, based on the replacement rules.

9. A computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams, received over a transport multiplex wherein multiple advertisements of different durations are being transmitted over a network, having (a) an ingest module, (b) at least one repository of content assets, (c) at least one repository of advertisement assets, (d) a playout server, (e) at least one repository of advertisement assets, and (f) at least one network streams, comprising the steps of:

a) Watermarking and Fingerprinting multiple advertisements A having audio-video streams, from a Master AV Database, placing an audio watermark on the incoming streams and extracting a high-confidence fingerprint (signature) from each of the watermarked assets A′;

b) communicating fingerprint signatures and replacement rules to at least one receiver over the Internet;

c) accessing at least one repository of local content assets;

d) receiving via at least one receiver at different locations, broadcast content including watermarked assets;

e) detecting advertisements having AV content, to be replaced, by matching their fingerprint signatures over a match duration using a signature detector; and

f) replacing detected advertisements, with local content assets, based on the replacement rules, at the at least one receiver,

wherein the at least one receiver has low complexity signature detectors further implementing the steps of:

a) receiving a broadcast AV stream for which it keeps generating the fingerprint for a window of audio, which has a fixed match duration;

b) matching the generated signature against all the references stored in the signature DB;

c) representing the signatures as a sequence of bit, a bit string;

d) detecting a match when the generated signature and a reference signature have the number of bits matching above a predefined threshold T; and

e) disambiguating at least two similar sounding master AVs having a match more than T by considering the master AV stream having the maximum matched bits compared to the current audio snapshot signature as the right master AV stream match.

10. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the step of Watermarking and Fingerprinting further comprises placing an audio watermark on the Master AV stream, the audio watermark being a pseudo-random, inaudible noise sequence.

11. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the advertisement assets from the Master AV Database have similar audio portions over a certain duration of their content.

12. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the step of Watermarking and Fingerprinting further comprises extracting a fingerprint signature from the watermarked content, the fingerprint being unique across assets with similar audio patterns toward the beginning due to the pseudo-random watermark thereby disambiguating assets with similar sounding audio over an initial match duration at the receiver.

13. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the at least one receiver has signature detectors further implementing the steps of:

a) performing watermark verification by extracting residual noise audio from incoming audio;

b) correlating the extracted noise against the pseudo-random, inaudible noise sequence generated by the WMFP module; and

c) detecting a match when the Master AV and the watermarked output match.

14. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the receivers at different locations monitor the transmitted content, actively replacing detected advertisements (A′) using the signature detector, with local content assets, based on the replacement rules.

15. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the replacement rules specify an identifier for the Master AV stream, the corresponding fingerprint signature and the replacement schedule on specific local receivers.

16. The computer-implemented method for low-latency detection of known audio-video content in multiple audio-video streams of claim 9 , wherein the at least one repository of local content assets comprise content assets that are specific to certain demographics, geographies, and customers.