PTAB

IPR2015-00348

Google Inc. v. Network-1 Technologies, Inc.

Title: System for Automatic Content Recognition and Action
Brief Description: The ’441 patent discloses a system for identifying an unknown electronic media work and performing a corresponding action based on the determined identity. The system operates by extracting features from the media work and using a non-exhaustive neighbor search to compare them against a database of reference works linked to action information.

Prior Art Relied Upon: Levy (Patent 6,505,160) and Arya (a 1998 publication titled "An Optimal Algorithm for Approximate Nearest Neighbor Searching in Fixed Dimensions").
Core Argument for this Ground:
- Prior Art Mapping: Petitioner argued that Levy discloses a system for identifying media content by extracting a "fingerprint" (i.e., features) and mapping it to a related action in a database. Arya was cited as disclosing an efficient, non-exhaustive neighbor search algorithm specifically designed to find "approximate nearest neighbor[s]" in multimedia databases, an improvement over computationally intensive brute-force searches.
- Motivation to Combine: A person of ordinary skill in the art (POSITA) would combine Levy with Arya to improve the efficiency and practicality of Levy's system. Because Levy's fingerprinting method could result in slight discrepancies between a source work and a captured sample, a search for near matches (as taught by Arya) would be a natural and necessary improvement to locate the correct content reliably.
- Expectation of Success: A POSITA would have a reasonable expectation of success because Arya was explicitly designed for the type of "multimedia databases" used in a system like Levy's, making its application straightforward.

Prior Art Relied Upon: Conwell (Patent 6,970,886).
Core Argument for this Ground:
- Prior Art Mapping: Petitioner asserted that Conwell discloses every element of the challenged claims. Conwell described a system that generates an identifier from digital content (extracting features) and looks it up in a database to retrieve a corresponding URL (the action). Petitioner argued that Conwell’s use of a hashing algorithm, which maps "similar or related, but non-identical, inputs" to the same output, constitutes the claimed "non-exhaustive neighbor search."
- Key Aspects: The core of this ground rested on construing Conwell’s single lookup in a sorted table using a hash identifier as an anticipatory non-exhaustive neighbor search, as it finds "neighbor" results without a brute-force comparison of all possible hashes.

Prior Art Relied Upon: Ghias (Patent 5,874,686) and Philyaw (Patent 6,098,106).
Core Argument for this Ground:
- Prior Art Mapping: Petitioner contended that Ghias teaches a query-by-humming system that identifies a melody by extracting its "contour representation" (features) and searching a database using an "efficient approximate pattern matching algorithm" (a non-exhaustive neighbor search) to return a list of matches. Philyaw was presented as teaching a system that links content to an action, specifically by extracting an embedded identifier from a broadcast to trigger a related advertisement.
- Motivation to Combine: A POSITA would have been motivated to combine the intrinsic content identification of Ghias with the advertising functionality of Philyaw. This combination would create a system capable of presenting content-relevant ads without the "time consuming and expensive step of embedding identifying information" required by Philyaw alone.
- Expectation of Success: The combination represented joining two known systems—one for content identification and one for advertising—to achieve a predictable and more efficient result.
Additional Grounds: Petitioner asserted an additional anticipation challenge under §102 against claims 1-3, 8-14, 19, 21-22, 25, 27, and 29-30 based on Wood (Patent 7,743,092).

"extracted features," "compact electronic representation," and "compact electronic representation comprising an extracted feature vector": Petitioner argued these terms are synonymous and should be construed broadly as "a quantitative representation of the features of a media work, or a subset of the features of a media work." This construction was asserted to be consistent with the specification and necessary to encompass the various fingerprinting and feature extraction techniques disclosed in the prior art.
"non-exhaustive ... search": Petitioner proposed construing this term to mean "a search that locates a match without conducting a brute force comparison of all possible matches, and all data within all possible matches." This broad definition was central to arguments that prior art hashing algorithms and approximate matching searches, which do not guarantee finding a match, meet this limitation.
"neighbor search": Consistent with the above, Petitioner argued this term should mean "identifying a close, but not necessarily exact, match," which was asserted to be the common understanding in the art at the time.

Petitioner requests institution of an inter partes review and cancellation of claims 1-3, 6-14, 18-19, 21-27, and 29-30 of the ’441 patent as unpatentable.