Detecting repeating content in broadcast media

1. A method executed by one or more computing devices, the method comprising:

determining a current match between an audio descriptor and a first reference descriptor, the current match occurring at a current time step and being associated with a current confidence score, the current time step having a time step length;

performing post-match processing, including:

determining a previous match between the audio descriptor and a second reference descriptor, the previous match occurring at a previous time step and being associated with a previous confidence score, the previous time step having the time step length; and

applying the following formula, upon condition that C h −1/L>C 0 , to the previous match and the previous confidence score:

{ M 0 ,C 0 }={M h ,C h −l/L}

where:

M 0 is a match for selecting content related to the audio descriptor;

C 0 is the current confidence score;

M h is the previous match occurring at the previous time step;

C h is the previous confidence score associated with the previous match;

l is a time step length; and

L is an expected dwell time;

selecting content related to the audio descriptor based on the match M 0 ; and

removing the selected content from a presentation to a user.

2. The method of claim 1 , where the first reference descriptor and second reference descriptor are frame descriptors.

3. The method of claim 1 , where the current match is a non-identity match.

4. The method of claim 1 , where determining the current match includes:

generating a short list of candidate auditory matches using hashing techniques; and

validating the candidate auditory matches using a validation procedure.

5. The method of claim 1 , where at least one of the previous confidence score or the current confidence score is a log likelihood confidence score given by an audio fingerprinting process.

6. A system, comprising:

a memory;

one or more processors configured to perform operations comprising:

determining a current match between an audio descriptor and a first reference descriptor, the current match occurring at a current time step and being associated with a current confidence score, the current time step having a time step length;

performing post-match processing, including:

determining a previous match between the audio descriptor and a second reference descriptor, the previous match occurring at a previous time step and being associated with a previous confidence score, the previous time step having the time step length; and

applying the following formula, upon condition that C h −1/L>C 0 , to the previous match and the previous confidence score:

{ M 0 ,C 0 }={M h ,C h −l/L}

where:

M 0 is a match for selecting content related to the audio descriptor;

C 0 is a the current confidence score;

M h is the previous match occurring at the previous time step;

C h is the previous confidence score associated with the previous match;

l is a time step length; and

L is an expected dwell time;

selecting content related to the audio descriptor based on the match M 0 ; and

removing the selected content from a presentation to a user.

7. The system of claim 6 , the operations further comprising:

validating the current match using non-auditory information.

8. The system of claim 6 , the operations further comprising:

determining end points of the repeating content.

9. The system of claim 8 , where the end points are determined using dynamic programming techniques.

10. The system of claim 6 , the operations further comprising:

applying metrics to the selected content to determine if the repeating content is an advertisement.

11. The system of claim 10 , where the metrics are from a group of metrics consisting of time duration, volume, visual activity, and blank frame bracketing.

12. The system of claim 6 , where the first reference descriptor and second reference descriptor are generated from ambient audio snippets of a media broadcast.

13. The system of claim 6 , where the first reference descriptor and second reference descriptor are frame descriptors.

14. The system of claim 6 , where the current match is a non-identity match.

15. The system of claim 14 , where the non-identity match is determined using hashing techniques.

16. The system of claim 6 , where at least one of the previous confidence score or the current confidence score is a log likelihood confidence score given by an audio fingerprinting process.

17. A non-transitory computer-readable storage medium having instructions stored thereon, which, when executed by a processor, causes the processor to perform operations comprising:

determining a current match between an audio descriptor and a first reference descriptor, the current match occurring at a current time step and being associated with a current confidence score, the current time step having a time step length;

performing post-match processing, including:

determining a previous match between the audio descriptor and a second reference descriptor, the previous match occurring at a previous time step and being associated with a previous confidence score, the previous time step having the time step length; and

applying the following formula, upon condition that C h −1/L>C 0 , to the previous match and the previous confidence score:

{ M 0 ,C 0 }={M h ,C h −l/L}

where:

M 0 is a match for selecting content related to the audio descriptor;

C 0 is the current confidence score;

M h is the previous match occurring at the previous time step;

C h is the previous confidence score associated with the previous match;

l is a time step length; and

L is an expected dwell time;

selecting content related to the audio descriptor based on the match M 0 ; and

providing the selected content for a presentation to a user.

18. The medium of claim 17 , the operations further comprising:

applying at least one metric to the selected content to determine if the repeating content is an advertisement.

19. The medium of claim 18 , wherein the metric is associated with one of a length of a media broadcast and a volume of a media broadcast.

20. The medium of claim 17 , where at least one of the previous confidence score or the current confidence score is a log likelihood confidence score given by an audio fingerprinting process.

21. A method executed by a computer, the method comprising:

determining a current match between an audio descriptor and a first reference descriptor, the current match occurring at a current time step and being associated with a current confidence score, the current time step having a time step length;

determining a previous match between the audio descriptor and a second reference descriptor, the previous match occurring at a previous time step and being associated with a previous confidence score, the previous time step having the time step length;

discounting the previous confidence score by a discount value, the discount value being calculated based on a ratio between the time step length and an expected dwell time;

determining that the discounted previous confidence score is greater than the current confidence score;

selecting content related to the audio descriptor based on the previous match; and

removing the selected content from a presentation to a user.

22. The method of claim 21 , where the current match is a non-identity match.

23. The method of claim 21 , where the expected dwell time indicates an expected time between channel changes.

24. The method of claim 21 , wherein the previous match is a best match of the previous time step.

25. The method of claim 24 , wherein each of the current confidence score and the previous confidence score includes a log-likelihood confidence score given by an audio fingerprinting process.

26. A system, comprising:

a processor; a memory;

one or more computers configured to perform operations comprising:

determining a current match between an audio descriptor and a first reference descriptor, the current match occurring at a current time step and being associated with a current confidence score, the current time step having a time step length;

determining a previous match between the audio descriptor and a second reference descriptor, the previous match occurring at a previous time step and being associated with a previous confidence score, the previous time step having the time step length;

discounting the previous confidence score by a discount value, the discount value being calculated based on a ratio between the time step length and an expected dwell time;

determining that the discounted previous confidence score is greater than the current confidence score;

selecting content related to the audio descriptor based on the previous match; and

removing the selected content from a presentation to a user.

27. The system of claim 26 , wherein the current match is a non-identity match.

28. The system of claim 26 , wherein the expected dwell time indicates an expected time between channel changes.

29. The system of claim 26 , wherein the previous match is a best match of the previous time step.

30. The system of claim 29 , wherein each of the current confidence score and the previous confidence score includes a log-likelihood confidence score given by an audio fingerprinting process.

31. A computer program product tangibly stored on a non-transitory computer-readable storage medium and operable to cause one or more computers to perform operations comprising:

determining a current match between an audio descriptor and a first reference descriptor, the current match occurring at a current time step and being associated with a current confidence score, the current time step having a time step length;

determining a previous match between the audio descriptor and a second reference descriptor, the previous match occurring at a previous time step and being associated with a previous confidence score, the previous time step having the time step length;

discounting the previous confidence score by a discount value, the discount value being calculated based on a ratio between the time step length and an expected dwell time;

determining that the discounted previous confidence score is greater than the current confidence score;

selecting content related to the audio descriptor based on the previous match; and

removing the selected content from a presentation to a user.

32. The product of claim 31 , wherein the current match is a non-identity match.

33. The product of claim 31 , wherein the expected dwell time indicates an expected time between channel changes.

34. The product of claim 31 , wherein the previous match is a best match of the previous time step.

35. The product of claim 34 , wherein each of the current confidence score and the previous confidence score includes a log-likelihood confidence score given by an audio fingerprinting process.