Signature generation for multimedia deep-content-classification by a large-scale matching system and method thereof

Patent metadata

Field	Value
Publication number	US-8990125-B2
Application number	US-201213682132-A
Country	US
Kind code	B2
Filing date	Nov 20, 2012
Priority date	Oct 26, 2005
Publication date	Mar 24, 2015
Grant date	Mar 24, 2015

Abstract

Content-based clustering, recognition, classification and search of high volumes of multimedia data in real-time. The embodiments disclosed herein are dedicated to real-time fast generation of signatures to high-volume of multimedia content-segments, based on relevant audio and visual signals, and to scalable matching of signatures of high-volume database of content-segments' signatures. The embodiments disclosed herein can be implemented in any applications which involve large-scale content-based clustering, recognition and classification of multimedia data, such as, content-tracking, video filtering, multimedia taxonomy generation, video fingerprinting, speech-to-text, audio classification, object recognition, video search and any other application requiring content-based signatures generation and matching for large content volumes such as, web and other large-scale databases.

First claim

What we claim is:

1. An apparatus for generating a signature for an input signal, comprising:
a processor; and
a memory coupled to the processor and configured to store at least instructions causing for execution of a plurality of computational cores, wherein the instructions includes a plurality of subsets of instructions, each subset of instructions causes the execution of one computational core from the plurality of computational cores and is coded to have properties that have at least some statistical independency from other subset of instructions, thereby the properties are independent of each other computational core, wherein each subset of instructions causes its respective computational core to generate responsive of the input signal a first signature element and a second signature element, the first signature element is a robust signature;
wherein a plurality of first signature elements include a first signature of the input signal and a plurality of second signature elements include a second signature of the input signal, the first signature is determined respective of a first threshold value and the second signature is determined respective of a second threshold, wherein the first threshold is higher than the second threshold.

2. The apparatus of claim 1 , wherein the first signature is robust to at least one of: noise and distortion.

3. The apparatus of claim 1 , wherein the properties include in part a first threshold and a second threshold, each of the plurality of computational cores generates responsive to the input signal and the first threshold the first signature element and responsive to the input signal and the second threshold the second signature element.

4. The apparatus of claim 3 , wherein each of the first threshold and the second threshold is a constant threshold value of a Heaviside step function, wherein each of the subset of instructions are coded with the Heaviside step function to be computed by each respective computational core when generating the first signature element and the second signature element.

5. The apparatus of claim 3 , wherein the properties are defined according to at least one random parameter.

6. The apparatus of claim 1 , wherein the input signal is at least multimedia signal.

7. The apparatus of claim 6 , wherein the multimedia signal is at least one of: an image, graphics, a video stream, a video clip, an audio stream, an audio clip, a video frame, text, a photograph, images of signals, and portions thereof.

8. A method for large-scale classification of multimedia signals, comprising:
generating by a signature generator a set of robust signatures representing samples for at least one class of multimedia signals;
for each multimedia signal to be classified performing:
generating at least a signature by the signature generator;
determining if the multimedia signal matches the at least one of class of multimedia signals based on the signature generated for the multimedia signal and the set of robust signatures representing samples for the at least one of class of multimedia signals; and
classifying the multimedia signal to a class of the multimedia signal matching the at least one of class of multimedia signals.

9. The method of claim 8 , further comprising:
creating a new class of multimedia signals to include an unmatched multimedia signal if is a match does not exist.

10. The method of claim 8 , wherein each robust signature of the set of robust signatures is robust to noise and distortion.

11. The method of claim 8 , wherein robust signatures of the set of robust signatures and the signature of the multimedia signal are generated by a signature generator.

12. The method of claim 11 , wherein the signature generator includes:
a processor; and
a memory coupled to the processor and configured to store at least instructions causing for execution of a plurality of computational cores, wherein the instructions includes a plurality of subsets of instructions, each subset of instructions causes the execution of one computational core from the plurality of computational cores and is coded to have properties that have at least some statistical independency from other subset of instructions, thereby the properties are independent of each other computational core, wherein each subset of instructions causes its respective computational core to generate responsive of a multimedia signal a first signature element and a second signature element, the first signature element is a robust signature;
wherein a plurality of first signature elements include a first signature of the input signal and a plurality of second signature elements include a second signature of the input signal, the first signature is determined respective of a first threshold value and the second signature is determined respective of a second threshold, wherein the first threshold is higher than the second threshold.

13. The method of claim 8 , wherein the multimedia signal is at least one of: an image, graphics, a video stream, a video clip, an audio stream, an audio clip, a video frame, text, a photograph, images of signals, and portions thereof.

Assignees

• Cortica Ltd

Inventors

• Raichelgauz Igal
• Odinaev Karina
• Zeevi Yehoshua Y

Classifications

• G06F16/1748

  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 title

• G06F16/2228

  Indexing structures · CPC title

• G06F16/48

  Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually · CPC title

• G06V20/48

  Matching video sequences · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title