Flagging advertisement frames for automatic content recognition

1. A method comprising:

receiving, at data processing hardware, broadcast fingerprints indicative of broadcast frames of a media stream, the media stream comprising a series of broadcast scenes;

receiving, at the data processing hardware, advertisement (ad) fingerprints indicative of ad frames of ad scenes;

determining, by the data processing hardware, an extent of similarity between (i) a last broadcast fingerprint of a last broadcast frame of a first broadcast scene and an (ii) initial broadcast fingerprint of an initial broadcast frame of a second broadcast scene;

determining, by the data processing hardware, a scene change between the first broadcast scene and the second broadcast scene based on the determined extent of similarity;

determining, by the data processing hardware, whether the second broadcast scene is one of the ad scenes; and

when the second broadcast scene is one of the ad scenes, associating, by the data processing hardware, an identification of the second broadcast scene as the one of the ad scenes.

2. The method of claim 1 , wherein:

determining the extent of similarity comprises determining, by the data processing hardware, a scene-change Pearson correlation coefficient between (i) the initial broadcast fingerprint of the initial broadcast frame of the second broadcast scene and (ii) the last broadcast fingerprint of the last broadcast frame of the first broadcast scene; and

determining the scene change comprises determining, by the data processing hardware, that the scene-change Pearson correlation coefficient satisfies a scene-change correlation threshold.

3. The method of claim 1 , wherein:

determining the extent of similarity comprises, for each sequentially adjacent pair of broadcast frames, determining a scene-change Pearson correlation coefficient between:

(i) a first broadcast fingerprint of a first broadcast frame of the pair of broadcast frames, and

(ii) a second broadcast fingerprint of a second broadcast frame of the pair of broadcast frames; and

determining the scene change comprises; for each sequentially adjacent pair of broadcast frames:

(i) determining, by the data processing hardware, whether the scene-change Pearson correlation coefficient satisfies a scene-change correlation threshold; and

(ii) when the scene-change Pearson correlation coefficient satisfies the scene-change correlation threshold:

(a) identifying the first broadcast frame of the pair of broadcast frames as a last broadcast frame of a corresponding broadcast scene; and

(b) identifying the second broadcast frame of the pair of broadcast frames as an initial broadcast frame of a corresponding sequentially adjacent broadcast scene.

4. The method of claim 1 , further comprising waiting a threshold time period after the last broadcast fingerprint of the last broadcast frame of the first broadcast scene before determining whether the second broadcast scene is one of the ad scenes.

5. The method of claim 1 , wherein determining whether the second broadcast scene is one of the ad scenes comprises:

for each ad fingerprint:

determining, by the data processing hardware, a match Pearson correlation coefficient between the respective ad fingerprint and the second broadcast fingerprint;

determining, by the data processing hardware, whether the match Pearson correlation coefficient satisfies a match correlation threshold; and

when the match Pearson correlation coefficient satisfies the match correlation threshold, identifying, by the data processing hardware, the second broadcast scene as corresponding to the ad scene of the respective ad fingerprint.

6. The method of claim 1 , further comprising, when the second broadcast scene is one of the ad scenes, determining, by the data processing hardware, whether the second broadcast scene is within or sequentially adjacent another ad scene.

7. The method of claim 1 , further comprising, when the second broadcast scene is one of the ad scenes, associating, by the data processing hardware, an ad identifier with the second broadcast scene.

8. The method of claim 1 , wherein each fingerprint represents at least one pixel of the corresponding frame.

9. The method of claim 1 , wherein each fingerprint comprises an average pixel value representing a sum of grayscale values of the corresponding frame.

10. The method of claim 1 , wherein each fingerprint represents a 16-integer vector corresponding to average pixel values of sixteen sub-frames, the sixteen sub-frames defining a four by four array of the corresponding frame.

11. A method comprising:

receiving, at data processing hardware, broadcast fingerprints indicative of broadcast frames of a media stream;

receiving, at the data processing hardware, advertisement (ad) fingerprints indicative of ad frames of ad scenes; and

for each sequentially adjacent pair of broadcast frames:

(i) determining, by the data processing hardware, a first extent of similarity between: (a) a first broadcast fingerprint of a first broadcast frame of the pair of broadcast frames, and (b) a second broadcast fingerprint of a second broadcast frame of the pair of broadcast frames;

(ii) determining, by the data processing hardware, whether the first extent of similarity is below a first similarity threshold; and

(iii) when the extent of similarity is below the similarity threshold:

(a) identifying, by the data processing hardware, a first broadcast scene as ending at the first broadcast frame;

(b) identifying, by the data processing hardware, a second broadcast scene as starting at the second broadcast frame; and

(c) for each ad fingerprint: (1) determining, by the data processing hardware, a second extent of similarity between the respective ad fingerprint and the second broadcast fingerprint, (2) determining, by the data processing hardware, whether the second extent of similarity is above a second similarity threshold, and (3) when the second extent of similarity is above the second similarity threshold, identifying, by the data processing hardware, the second broadcast scene as the ad scene of the respective ad fingerprint.

12. The method of claim 11 , further comprising, when the second broadcast scene is one of the ad scenes, determining, by the data processing hardware; whether the second broadcast scene is within or sequentially adjacent another ad scene.

13. The method of claim 11 , further comprising, when the second broadcast scene is one of the ad scenes, associating, by the data processing hardware, an ad identifier with the second broadcast scene.

14. The method of claim 11 , wherein each fingerprint represents at least one pixel of the corresponding frame.

15. The method of claim 11 , wherein each fingerprint represents a 16-integer vector corresponding to average pixel values of sixteen sub-frames, the sixteen sub-frames defining a four by four array of the corresponding frame.

16. A system comprising:

data processing hardware; and

memory hardware in communication with the data processing hardware, the memory hardware storing instructions that when executed on the data processing hardware cause the data processing hardware to perform operations comprising:

receiving broadcast fingerprints indicative of broadcast frames of a media stream, the media stream comprising a series of broadcast scenes;

receiving advertisement (ad) fingerprints indicative of ad frames of ad scenes;

determining an extent of similarity between (i) a last broadcast fingerprint of a last broadcast frame of a first broadcast scene and an (ii) initial broadcast fingerprint of an initial broadcast frame of a second broadcast scene;

determining a scene change between the first broadcast scene and the second broadcast scene based on the determined extent of similarity;

determining whether the second broadcast scene is one of the ad scenes; and

when the second broadcast scene is one of the ad scenes, associating an identification of the second broadcast scene as the one of the ad scenes.

17. The system of claim 16 , wherein:

determining the extent of similarity comprises determining a scene-change Pearson correlation coefficient between (i) the initial broadcast fingerprint of the initial broadcast frame of the second broadcast scene and (ii) the last broadcast fingerprint of the last broadcast frame of the first broadcast scene; and

determining the scene change comprises determining that the scene-change Pearson correlation coefficient satisfies a scene-change correlation threshold.

18. The system of claim 16 , wherein:

determining the extent of similarity comprises, for each sequentially adjacent pair of broadcast frames, determining a scene-change Pearson correlation coefficient between:

(i) a first broadcast fingerprint of a first broadcast frame of the pair of broadcast frames, and

a second broadcast fingerprint of a second broadcast frame of the pair of broadcast frames; and

determining the scene change comprises, for each sequentially adjacent pair of broadcast frames:

(i) determining whether the scene-change Pearson correlation coefficient satisfies a scene-change correlation threshold; and

(ii) when the scene-change Pearson correlation coefficient satisfies the scene-change correlation threshold:

(a) identifying the first broadcast frame of the pair of broadcast frames as a last broadcast frame of a corresponding broadcast scene; and

(b) identifying the second broadcast frame of the pair of broadcast frames as an initial broadcast frame of a corresponding sequentially adjacent broadcast scene.

19. The system of claim 16 , wherein the operations further comprise waiting a threshold time period after the last broadcast fingerprint of the last broadcast frame of the first broadcast scene before determining whether the second broadcast scene is one of the ad scenes.

20. The system of claim 16 , wherein determining whether the second broadcast scene is one of the ad scenes comprises:

for each ad fingerprint:

determining a match Pearson correlation coefficient between the respective ad fingerprint and the second broadcast fingerprint;

determining whether the match Pearson correlation coefficient satisfies a match correlation threshold; and

when the match Pearson correlation coefficient satisfies the match correlation threshold, identifying the second broadcast scene as corresponding to the ad scene of the respective ad fingerprint.

21. The system of claim 16 , wherein the operations further comprise, when the second broadcast scene is one of the ad scenes, determining whether the second broadcast scene is within or sequentially adjacent another ad scene.

22. The system of claim 16 , wherein the operations further comprise, when the second broadcast scene is one of the ad scenes, associating an ad identifier with the second broadcast scene.

23. The system of claim 16 , wherein each fingerprint represents at least one pixel of the corresponding frame.

24. The system of claim 16 , wherein each fingerprint comprises an average pixel value representing a sum of grayscale values of the corresponding frame.

25. The system of claim 16 , wherein each fingerprint represents a 16-integer vector corresponding to average pixel values of sixteen sub-frames, the sixteen sub-frames defining a four by four array of the corresponding frame.

26. A system comprising:

data processing hardware; and

memory hardware in communication with the data processing hardware, the memory hardware storing instructions that when executed on the data processing hardware cause the data processing hardware to perform operations comprising:

receiving broadcast fingerprints indicative of broadcast frames of a media stream;

receiving advertisement (ad) fingerprints indicative of ad frames of ad scenes; and

for each sequentially adjacent pair of broadcast frames:

(i) determining, by the data processing hardware, a first extent of similarity between: (a) a first broadcast fingerprint of a first broadcast frame of the pair of broadcast frames, and (b) a second broadcast fingerprint of a second broadcast frame of the pair of broadcast frames;

(ii) determining, by the data processing hardware, whether the first extent of similarity is below a first similarity threshold; and

(iii) when the extent of similarity is below the similarity threshold:

(a) identifying, by the data processing hardware, a first broadcast scene as ending at the first broadcast frame;

(b) identifying, by the data processing hardware, a second broadcast scene as starting at the second broadcast frame; and

(c) for each ad fingerprint: (1) determining, by the data processing hardware, a second extent of similarity between the respective ad fingerprint and the second broadcast fingerprint, (2) determining, by the data processing hardware, whether the second extent of similarity is above a second similarity threshold, and (3) when the second extent of similarity is above the second similarity threshold, identifying, by the data processing hardware, the second broadcast scene as the ad scene of the respective ad fingerprint.

27. The system of claim 26 , wherein the operations further comprise, when the second broadcast scene is one of the ad scenes, determining whether the second broadcast scene is within or sequentially adjacent another ad scene.

28. The system of claim 26 , wherein the operations further comprise, when the second broadcast scene is one of the ad scenes, associating an ad identifier with the second broadcast scene.

29. The system of claim 26 , wherein each fingerprint represents at least one pixel of the corresponding frame.

30. The system of claim 26 , wherein each fingerprint represents a 16-integer vector corresponding to average pixel values of sixteen sub-frames, the sixteen sub-frames defining a four by four array of the corresponding frame.