PTAB

IPR2020-00219

Free Stream Media Corp v. Gracenote Inc

Log In
Key Events
Petition

1. Case Identification

2. Patent Overview

  • Title: Method and Apparatus for Multi-Dimensional Content Search and Video Identification
  • Brief Description: The ’030 patent discloses a system for identifying video content by searching a database. The system generates a "robust hash" video signature from a query video and uses it as a traversal index to efficiently search a multidimensional, hierarchical database containing reference signatures.

3. Grounds for Unpatentability

Ground 1: Claims 1, 8, 11, 13, and 15 are obvious over Satoh and Roover.

  • Prior Art Relied Upon: Satoh (“Indexing Video Archives,” 2004) and Roover (“Robust Video Hashing Based on Radial Projection of Key Frames,” Oct. 2005).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Satoh disclosed the core elements of claim 1, including a method for organizing a multidimensional video database using a hierarchical, tree-like structure (e.g., an R-tree) with root, intermediate, and leaf nodes. Satoh taught generating high-dimensional "feature vectors" from video frames (e.g., based on color information from an extracted face) and using these vectors as a "traversal index" to search the database and retrieve data from the leaf nodes. While Satoh did not explicitly teach hashing its vectors, Petitioner contended that Roover remedied this by teaching a "robust image hashing algorithm" specifically for video sequences. Roover explained that applying a robust hash to feature vectors makes them resilient to common distortions like compression or scaling.
    • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine these references to improve the performance of Satoh’s video search system. Roover expressly taught that robust hashing is an "obvious solution" for content identification and indexing and can be combined with "conventional digital signature methods" like those in Satoh. Applying Roover's robust hash would make Satoh's search system less susceptible to common video manipulations, enhancing its ability to return relevant search results. Hashing also provides the known benefits of reducing data size and improving search speed.
    • Expectation of Success: A POSITA would have a reasonable expectation of success because both references address the same technical problem of video identification using multidimensional vectors. Roover’s disclosure that its hashing algorithm could be applied to various "more complex vectors" would confirm to a POSITA that it could be successfully applied to the high-dimensional feature vectors taught by Satoh.

Ground 2: Claims 1, 6, 8, 10, 11, 13, and 15 are obvious over Deng and Brunk.

  • Prior Art Relied Upon: Deng (WO 2005/036877A1) and Brunk (Application # 2002/0126872).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Deng disclosed nearly all limitations, including a system for identifying video using a hierarchical, "linked tree-like data structure." Deng taught generating multidimensional signatures from video frames based on multiple parameters, such as the centroids of color and brightness components. These signatures were designed to be robust against scaling and shifting and were used as an index to traverse the database to a "final table" (leaf node) containing associated metadata. Petitioner asserted that Brunk supplied the final element by teaching that content signatures can and should be subject to a "hash" to produce a unique, robust identifier. Brunk further taught that a hash could be based on shape features, such as the center of mass relative to an object’s edges, mapping to dependent claim 6.
    • Motivation to Combine: A POSITA would combine Deng and Brunk to gain the well-known benefits of hashing. Deng’s system was designed for fast searches and transmission of signatures over a network. Brunk taught that hashing reduces signature size and improves "database efficiency" and access speed. A POSITA would have been motivated to apply Brunk's hashing method to Deng's already robust signatures to further reduce their size, which would improve storage efficiency, decrease network bandwidth requirements for transmission, and increase search speed in Deng’s database.
    • Expectation of Success: A POSITA would expect success in combining these references because both relate to generating robust content signatures from video. Brunk taught that its hashing algorithm can be applied to signatures generated from a wide range of features, including the same "center of mass" features used by Deng. This direct technical overlap would give a POSITA every reason to believe that Brunk's hashing algorithm could be successfully applied to Deng's signatures.

4. Key Claim Construction Positions

  • "Robust hash": Petitioner proposed the construction "a corresponding smaller numerical value calculated from another value such that it is more immune to noise and disturbances due to distortion." This construction was argued to be consistent with the plain meaning and the patent’s description of a hash that is "more immune to noise and disturbances."
  • "Traversal index": Petitioner proposed the construction "an index for stepping or traversing through the nodes of a hierarchical database." This was based on the patent's disclosure of using the index to navigate from a root node to leaf nodes in a tree-like structure.
  • "Leaf node": Petitioner proposed the construction "a node in a hierarchical data structure without any children." This was argued to align with the plain and ordinary meaning and the patent’s depiction of leaf nodes at the bottom of the database hierarchy.

5. Relief Requested

  • Petitioner requested institution of an inter partes review and cancellation of claims 1, 6, 8, 10, 11, 13, and 15 of the ’030 patent as unpatentable.