SYSTEM AND METHODS FOR GENERATION OF A CONCEPT BASED DATABASE

• US 20130159311A1
• Filed: 02/13/2013
• Published: 06/20/2013
• Est. Priority Date: 10/26/2005
• Status: Active Grant

• Current Assignee: Cortica Ltd.
• Original Assignee: Cortica Ltd.
• Inventors: Raichelgauz, Igal, Ordinaev, Karina, Zeevi, Yehoshua Y.
• Granted Patent: US 9,031,999 B2
• Time in Patent Office: Days
• US Class Current: 707/737
• CPC Class Codes: G06F 16/1748 De-duplication implemented within the file system, e.g. based on file segments de-duplication techniques in storage systems for the management of data blocks G06F3/0641
• CPC Class Codes: G06F 16/2228 Indexing structures
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/41 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/43 Querying
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface
• CPC Class Codes: Y10S 707/99948 Application of database or data structure, e.g. distributed, multimedia, or image

Abstract

A method for generating a concept database respective of a plurality of multimedia data elements (MMDEs) comprises generating a plurality of items from a received MMDE of the plurality of MMDEs; determining the items that are of interest for signature generation; generating at least one signature responsive to at least one item of interest of the received MMDE of the plurality of MMDEs; clustering a plurality of signatures received from the signature generator responsive of the plurality of MMDEs; reducing the number of signatures in each cluster to a create a signature reduced cluster (SRC) of the cluster; associating metadata with the SRC to a concept structure comprised of a plurality of SRCs and their associated metadata; and generating at least one index for mapping the received MMDE to at least one concept structure, wherein the concept database includes concept structures and the generated indices for the plurality of MMDEs.

Claims

1. A system for generating a concept database respective of a plurality of multimedia data elements (MMDEs), comprising:

an attention processor (AP) for generating a plurality of items from a received MMDE of the plurality of MMDEs and determining which of the generated items are of interest for signature generation;

a signature generator (SG) for generating at least one signature responsive to at least one item of interest of the received MMDE of the plurality of MMDEs;

a clustering processor (CP) for clustering a plurality of signatures received from the signature generator responsive of the plurality of MMDEs and creating a signature reduced cluster (SRC);

a concept generator (CG) for associating metadata with the SRC and forming a concept structure comprised of a plurality of SRCs and their associated metadata; and

an index generator (IG) for generating at least one index for mapping the received MMDE to at least one concept structure, wherein the concept database includes concept structures and the generated indices for the plurality of MMDEs.

2. The system of claim 1, wherein the AP is a patch AP (PAP) that extracts items in the form of patches from the received MMDE.

3. The system of claim 2, wherein the PAP determines which patches to provide for the signature generation based on at least one of:

an entropy level of a patch, corners identified in a patch, and borders identified in a patch.

4. The system of claim 1, wherein the AP extracts items in the form of patterns from the received MMDE.

5. The system of claim 1, wherein the CP is configured to create the SRC of the cluster by:

reducing the number of signatures in each cluster to a minimum that still identifies all the MMDEs associated with the cluster.

6. The system of claim 1, wherein the MMDE is at least one of:

an image, graphics, a video stream, a video clip, an audio stream, an audio clip, a video frame, a photograph, images of signals, medical signals, geophysical signals, subsonic signals, supersonic signals, electromagnetic signals, and infrared signals.

7. The system of claim 1, wherein the SG comprises a plurality of computational cores, each computational core having properties statistically independent of each other computational core, each computational core generates responsive to the received MMDE at least one signature, wherein the at least one signature is robust to noise and distortions.

8. The system of claim 1, wherein the CP is further configured to:

generate a clustering score for each signature of an MMDE of the plurality of MMDEs versus all other MMDEs of the plurality of MMDEs;

determine a size of a diagonal matrix having a size corresponding to the number of the plurality of MMDEs;

place the clustering score in a diagonal matrix in storage, one clustering score for each pair of MMDEs;

create a new cluster element for each two cluster elements in the diagonal matrix having a clustering score that exceeds a threshold; and

repeat the process at the newly created cluster element level each time using the new cluster elements generated as the plurality of cluster elements for the subsequent sequence, until a single cluster is reached or it is determined that a single cluster cannot be reached.

9. The system of claim 1, wherein the associated metadata is at least one of:

a plurality of patches created by the PAP from each MMDE, the at least one signature generated by the SG respective of each MMDE, the concept matched for each MMDE respective of the signatures generated by the CP.

10. The system of claim 1, wherein the IG is further configured to:

match metadata related to the received MMDE against concept structures;

generate the index for the received MMDE, for each matching concept structure; and

store a plurality of indices generated for the plurality of MMDEs in the concept database.

11. A method for generating a concept database respective of a plurality of multimedia data elements (MMDEs) comprising:

generating a plurality of items from a received MMDE of the plurality of MMDEs;

determining the items that are of interest for signature generation;

generating at least one signature responsive to at least one item of interest of the received MMDE of the plurality of MMDEs;

clustering a plurality of signatures received from the signature generator responsive of the plurality of MMDEs;

reducing the number of signatures in each cluster to a create a signature reduced cluster (SRC) of the cluster;

associating metadata with the SRC to a concept structure comprised of a plurality of SRCs and their associated metadata; and

generating at least one index for mapping the received MMDE to at least one concept structure, wherein the concept database includes concept structures and the generated indices for the plurality of MMDEs.

12. The method of claim 11, wherein generating the plurality of items further comprising:

extracting items in the form of patches from the received MMDE by a patch AP (PAP).

13. The method of claim 11, further comprising:

determining which patches to provide for the signature generation based on at least one of;

an entropy level of a patch, corners identified in a patch, and borders identified in a patch.

14. The method of claim 11, wherein generating the plurality of items further comprising:

extracting items in the form of patterns from the received MMDE.

15. The method of claim 11, wherein creating the SRC further comprising:

reducing the number of signatures in each cluster to a minimum that still identifies all the MMDEs associated with the cluster.

16. The method of claim 11, wherein the MMDE is at least one of:

an image, graphics, a video stream, a video clip, an audio stream, an audio clip, a video frame, a photograph, images of signals, medical signals, geophysical signals, subsonic signals, supersonic signals, electromagnetic signals, and infrared signals.

17. The method of claim 11, wherein the SG comprises a plurality of computational cores, each computational core having properties statistically independent of each other computational core, each computational core generates, responsive to the received MMDE at least one signature, wherein the least one signature is robust to noise and distortions.

18. The method of claim 11, wherein clustering the plurality of signatures further comprising:

generating a clustering score for each signature of an MMDE of the plurality of MMDEs versus all other MMDEs of the plurality of MMDEs;

determining a size of a diagonal matrix having a size corresponding to the number of the plurality of MMDEs;

placing the clustering score in a diagonal matrix in storage, one clustering score for each pair of MMDEs;

creating a new cluster element for each two cluster elements in the diagonal matrix having a clustering score that exceeds a threshold; and

repeating the process at the newly created cluster element level each time using the new cluster elements generated as the plurality of cluster elements for the subsequent sequence, until a single cluster is reached or it is determined that a single cluster cannot be reached.

19. The method of claim 11, wherein the associated metadata is at least one of:

a plurality of patches from each MMDE, the at least one signature generated respective of each MMDE, the concept matched for each MMDE respective of the at least one signature.

20. The method of claim 11, generating the at least one index further comprising:

matching metadata related to the received MMDE against concept structures; and

generating the index for the received MMDE, for each matching concept structure; and

storing a plurality of indices generated for the plurality of MMDEs in the concept database.

21. A non-transitory computer readable medium having stored thereon instructions for causing one or more processing units to execute the method according to claim 11.