Apple Inc v. Andrea Electronics Corp

1. Case Identification

• Case #: IPR2017-00732
• Patent #: 6,377,637
• Filed: January 20, 2017
• Petitioner(s): Apple Inc.
• Patent Owner(s): Andrea Electronics Inc.
• Challenged Claims: 1-14

2. Patent Overview

• Title: Sub-Band Exponential Smoothing Noise Canceling System
• Brief Description: The ’637 patent discloses a system for reducing noise in a digital audio signal using a spectral subtraction process. The system splits an input signal into multiple sub-bands, processes each sub-band to cancel noise using techniques like exponential smoothing and adaptive thresholding, and then recombines the sub-bands into a clean output signal.

3. Grounds for Unpatentability

Ground 1: Claims 1-14 are obvious over Hirsch in view of Martin.

• Prior Art Relied Upon: Hirsch (a 1995 IEEE publication) and Martin (a 1994 EUSIPCO publication).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that independent claims 1 (apparatus) and 8 (method) are obvious over the combination of Hirsch and Martin. Hirsch was asserted to teach a complete noise reduction system that splits a signal into sub-bands and uses an "adaptive threshold" based on a running noise average to distinguish speech from noise. Martin was asserted to teach a complementary spectral subtraction method that estimates the noise floor by tracking the minimum signal value over a period and uses exponential smoothing of the sub-band signal power. The combination of Hirsch's adaptive threshold concept with Martin's minimum-statistics noise estimation and signal smoothing allegedly taught all key limitations of the independent claims.
  • Motivation to Combine: A Person of Ordinary Skill in the Art (POSITA) would combine Hirsch and Martin because Hirsch expressly stated its system could be combined with other "well known spectral subtraction techniques." Hirsch noted its own limitations in handling non-stationary noise and pointed to Martin's algorithm as a known solution to that exact problem. This provided a direct and explicit reason to integrate Martin's superior noise estimation into Hirsch's framework to improve performance in varied noise environments.
  • Expectation of Success: A POSITA would have had a reasonable expectation of success in combining the references. Integrating Martin's noise floor algorithm into Hirsch's system was presented as a simple substitution of one noise estimation module for another, an "optimization" that would predictably improve performance without altering the system's overall function.

Ground 2: Claims 1-14 are obvious over Hirsch and Martin in further view of Crochiere.

• Prior Art Relied Upon: Hirsch (a 1995 IEEE publication), Martin (a 1994 EUSIPCO publication), and Crochiere (a 1983 textbook on digital signal processing).
• Core Argument for this Ground:
  • Prior Art Mapping: This ground reinforced Ground 1, focusing on the specific claim requirement of using a Discrete Fourier Transform (DFT) filter bank with single sideband (SSB) modulation. Petitioner argued that while Hirsch and Martin taught using DFT filter banks, the Crochiere textbook provided the explicit, enabling details for the specific DFT-SSB type recited in the claims.
  • Motivation to Combine: The motivation was exceptionally strong because the Martin reference itself cited Crochiere for guidance on implementing DFT filter banks. Furthermore, the challenged ’637 patent also relied on Crochiere for the same purpose. Petitioner asserted that a POSITA implementing the Hirsch/Martin combination would have been led directly to Crochiere, which taught that the DFT-SSB variant was preferable for speech processing applications, making its selection obvious.

Ground 3: Claims 6-7 and 13-14 are obvious over Hirsch and Martin in further view of Ertem.

• Prior Art Relied Upon: Hirsch (a 1995 IEEE publication), Martin (a 1994 EUSIPCO publication), and Ertem (Patent 6,453,289).

• Core Argument for this Ground:

  • Prior Art Mapping: This ground targeted dependent claims adding two specific features to the subtraction process: performing a "minimum filter coefficient threshold function" (claims 6, 13) and "exponential smoothing of said filter coefficient" (claims 7, 14). Petitioner argued that while the base Hirsch/Martin combination taught the core noise subtraction, Ertem taught these specific enhancements. Ertem disclosed that setting a minimum floor for the filter coefficient and smoothing the coefficient over time were known techniques to reduce audible artifacts and improve the listening experience.
  • Motivation to Combine: A POSITA seeking to optimize the combined Hirsch/Martin system would have been motivated to reduce processing artifacts, a common problem in spectral subtraction. Ertem was presented as teaching solutions to this exact problem. Therefore, a POSITA would have been motivated to incorporate Ertem's techniques for setting a minimum filter threshold and smoothing the filter coefficient into the Hirsch/Martin system to enhance its performance and output quality.

• Additional Grounds: Petitioner asserted an additional obviousness challenge against claims 6-7 and 13-14 based on the combination of Hirsch, Martin, Crochiere, and Ertem, relying on similar arguments.

4. Key Claim Construction Positions

• Means-Plus-Function Terms: Petitioner dedicated significant argument to construing several means-plus-function limitations under §112(f).
• "band splitting means" / "recombining means": Petitioner argued that the ’637 patent disclosed no specific algorithm for these functions, instead citing the Crochiere textbook. Therefore, the corresponding structure must be construed as "a conventional DFT filter bank using single side band modulation technique as described in Crochiere." This construction was critical to show the claimed structure was taught by the prior art.
• "noise estimating means": Petitioner contended the corresponding structure for this term was the entire algorithm depicted in Figure 4 of the ’637 patent, including all its sub-components (future/current minimum calculation, thresholding, and exponential averaging). This construction was key to Petitioner's argument that the combination of Hirsch and Martin taught all components of the claimed structure.
• "subtraction processing means": The structure was argued to be the algorithm in Figure 5 for generating and applying a filter coefficient. For dependent claims 6 and 7, the structure further included optional elements from Figure 5 for applying a minimum limit and smoothing the coefficient.

5. Relief Requested

• Petitioner requested the institution of an inter partes review and the cancellation of claims 1-14 of the ’637 patent as unpatentable under 35 U.S.C. §103.