Apple Inc v. Qualcomm Inc

1. Case Identification

• Case #: IPR2018-01282
• Patent #: 8,768,865
• Filed: June 29, 2018
• Petitioner(s): Apple Inc.
• Patent Owner(s): Qualcomm Incorporated
• Challenged Claims: 1-10, 12-30, 46-53

2. Patent Overview

• Title: Learning Situations Via Pattern Matching
• Brief Description: The ’865 patent discloses a machine learning system for mobile devices to recognize user situations. The system aims to reduce computational load by first detecting a specific "condition" (e.g., a time of day or user action) and then initiating a more complex pattern recognition process only after that condition is met.

3. Grounds for Unpatentability

Ground 1: Anticipation of Claims 1-4, 15-17, 21-23, 28, 29, 46, and 47 under §102 by Louch

• Prior Art Relied Upon: Louch (Patent 8,676,224).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Louch, which discloses a system for automatically controlling a mobile device's speakerphone based on its current "state," teaches every limitation of the challenged independent claims. Louch allegedly discloses monitoring signals from multiple sensors (e.g., accelerometer, proximity sensor), detecting a "state" (which Petitioner mapped to the claimed "condition"), and learning or identifying "patterns of the state" (mapped to the claimed "first pattern"). Petitioner further contended Louch teaches fixing parameters by recording a first movement pattern in a learning mode and later comparing it to a second detected pattern, which satisfies the claim limitation of "fixing a subset of varying parameters...to represent said at least one detected condition."
  • Key Aspects: The argument centered on equating Louch's "state" detection and "pattern" learning with the ’865 patent's two-step condition-then-pattern recognition framework, asserting Louch’s system inherently performs these steps to control device functions.

Ground 2: Obviousness of Claims 5-10, 18-20, 24-27, 30, and 48-53 under §103 over Louch in view of Nadkarni

• Prior Art Relied Upon: Louch (Patent 8,676,224) and Nadkarni (Application # 2010/0217533).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner asserted that Louch provides the foundational system for state-based pattern recognition on a mobile device. Nadkarni, which teaches identifying human activities by matching accelerometer signals against a stored library of "signatures," was argued to supply the limitations missing from Louch, particularly regarding the elimination of irrelevant patterns. For instance, Nadkarni teaches filtering signals and using a "previous human activity context" to eliminate certain signatures from consideration, which Petitioner mapped to the claims requiring identification of an "irrelevant pattern" via a "context labeling-type process."
  • Motivation to Combine: A POSITA would combine Louch and Nadkarni to improve the efficiency and accuracy of Louch's system. Nadkarni's techniques for reducing the set of possible patterns by eliminating irrelevant ones based on context would have been a known method to solve the problem of computational complexity in pattern recognition systems like Louch's.
  • Expectation of Success: A POSITA would have a reasonable expectation of success because combining Nadkarni's filtering and context-based elimination methods with Louch's state-detection system involved applying known signal processing techniques to achieve the predictable result of a more efficient pattern matching process.

Ground 3: Obviousness of Claims 12-14 under §103 over Louch and Nadkarni in view of Greenhill

• Prior Art Relied Upon: Louch (Patent 8,676,224), Nadkarni (Application # 2010/0217533), and Greenhill (Application # 2008/0297513).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground specifically targeted the database-related claims. Petitioner argued the combination of Louch and Nadkarni teaches the core pattern recognition system. Greenhill, which discloses a data management system for industrial processes, was introduced for its explicit teaching of using a "condition database" (disclosed as an "event database" for storing alarm conditions) and a "correlation database" to store relationships between different data streams. These specific database structures, absent in Louch and Nadkarni, were argued to be taught by Greenhill.
  • Motivation to Combine: A POSITA implementing the pattern recognition system of Louch/Nadkarni would be motivated to incorporate Greenhill's structured database system to efficiently manage and store the sensor data, condition data, and pattern information. Using distinct databases for conditions and correlations was presented as a known and logical design choice for organizing data in a complex machine learning system to improve speed and reduce storage size.
  • Expectation of Success: Implementing Greenhill's database management scheme in Louch's system would be a straightforward application of known data structuring principles, and a POSITA would expect success without affecting the underlying functionality of the pattern recognition process.

4. Key Claim Construction Positions

• "condition": Petitioner argued this term should be construed broadly to encompass any "state or action," including device state, user actions (e.g., walking, driving), or time of day, as supported by the specification. This broad construction allowed Petitioner to map the "state" of the mobile device in Louch to the claimed "condition."
• "pattern": Petitioner proposed this term encompasses a "collection of one or more parameter values," such as sensor readings (e.g., "Loud" sound intensity, "Running" movement). This construction was crucial for mapping Louch's learned "patterns of the state," which are comprised of sensor data, to the claimed "pattern."
• "fixing a subset of varying parameters...": Petitioner argued this phrase is satisfied by "associating at least one parameter...with the first pattern to represent at least one detected condition," based on amendments made during prosecution. This construction allowed Petitioner to argue that Louch's act of recording a pattern (a collection of parameters) to represent a specific device state met this limitation.
• "condition database" and "correlation database": For claims 12-14, Petitioner argued these terms should be construed as databases for storing condition/event data and correlation data, respectively. This construction was key to the argument that Greenhill, which explicitly discloses an "event database" and "correlation database," rendered these claims obvious.

5. Relief Requested

• Petitioner requests institution of an inter partes review and cancellation of claims 1-10, 12-30, and 46-53 of the ’865 patent as unpatentable.