Multimedia stream analysis and retrieval

1. A method of processing a video stream to determine segment boundaries in the video stream, comprising:

performing a content analysis of the video stream, comprising:

for a plurality of parts of the video stream, determining a probability that each part represents the start of a new segment;

for the plurality of parts of the video stream, adjusting the probability in dependence on a proximity in time of each part to a last identified segment boundary;

identifying a transition from one segment to another segment in the video stream based on the content analysis; and

marking a segment boundary corresponding to the identified transition.

2. The method of claim 1 further comprising:

determining a content analysis to be applied to the video stream;

performing the content analysis of the video stream based on that determination.

3. The method of claim 2 wherein the content analysis includes a plurality of analysis modes, the method further comprising:

aggregating the analysis modes.

4. The method of claim 1 wherein the content analysis is at least one of a plurality of analysis modes.

5. The method of claim 4 wherein the plurality of analysis modes include: video analysis; image analysis; audio analysis; and textual analysis.

6. The method of claim 5 wherein identifying a transition from one segment to another segment comprises identifying a change in any one of video, image, audio, and text of the video stream.

7. The method of claim 4 wherein each analysis mode affects a probability that a given moment in the video stream indicates a start of a new segment.

8. The method of claim 1 , wherein the step of adjusting the probability in dependence on the proximity in time of each part to the last identified segment boundary comprises:

increasing the probability the further in time the part of the video stream is from the last identified segment boundary.

9. The method of claim 1 wherein identifying a transition from one segment to another segment comprises assuming the video stream comprises a linear sequence of segments.

10. The method of claim 1 wherein the content analysis includes a textual analysis, and wherein for the textual analysis identifying a transition from one segment to another segment comprises identifying a start of sentences, and examining the sentences.

11. The method of claim 10 wherein examining the sentences comprises comparing one sentence to another.

12. The method of claim 10 wherein examining the sentences comprises comparing each sentence to one of a previous sentence and a combination of previous sentences, in order to identify a transition from one segment to another segment.

13. The method of claim 10 , further comprising:

applying a score to an analysis of each sentence, the score denoting a likelihood that each sentence is associated with the start of a new segment.

14. The method of claim 13 , wherein the score is dependent upon a number of sentences between a current sentence and a start of the new segment, a greater number of sentences resulting in a higher probability that the current sentence is associated with a new segment.

15. The method of claim 13 further comprising:

modelling the scores in order to determine segment boundaries.

16. The method of claim 1 wherein the transition from one segment to another segment is based on identifying a transition from one theme to another theme.

17. The method of claim 16 wherein the content analysis determines whether a theme of the video stream changes, and in dependence on selecting a change in themes, a transition between segments is identified.

18. The method of claim 1 , further comprising:

applying the content analysis to an entirety of content of each segment; and

generating a descriptor for each segment providing a summary of the segment.

19. The method of claim 1 , further comprising:

identifying a topic of interest to a user;

searching a database of segments of content, the data base including a summary of topics of each segment;

comparing the topic of interest to the user to the topics of each segment;

providing recommendations to the user based on a match.

20. A device for processing a video stream to determine segment boundaries, comprising:

a content analysis module for performing a content analysis of the video stream stored and performed on a memory device, and configured to:

determine a probability, for a plurality of parts of the video stream, that each part represents the start of a new segment; and

adjust the determined probability, for each of the plurality of parts of the video stream, in dependence on a proximity in time of each part to a last identified segment boundary; and

a processing module stored and performed on the memory device and configured to identify a transition from one segment to another segment in the video stream based on the content analysis, and mark a segment boundary corresponding to the identified transition.