PTAB

IPR2025-01096

Amazon.com Inc v. SoundClear Technologies LLC

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Noise Reduction Apparatus
  • Brief Description: The ’259 patent relates to a system for detecting speech and reducing noise in audio signals captured by at least two microphones. The apparatus uses a "speech segment determiner" to identify speech, a "voice direction detector" to determine the direction of arrival (DOA) of the speech, and an "adaptive filter" that uses this information to perform a noise reduction process.

3. Grounds for Unpatentability

Ground 1: Obviousness over Chen and Li - Claims 1-3, 5-6, 8-9, 12-15, and 17-20 are obvious over Chen in view of Li.

  • Prior Art Relied Upon: Chen (Application # 2010/0081487) and Li (Application # 2012/0197638).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Chen discloses a mobile device that enhances voice quality using multiple microphones and a digital signal processor (DSP). This DSP meets the limitations for a "speech segment determiner" and a "voice direction detector" by analyzing microphone signals based on signal-to-noise ratio (SNR) and DOA to select a "voice-dominant" signal. Petitioner contended Chen’s "echo/noise cancellation controller" performs the claimed noise reduction but does not explicitly name it an "adaptive filter." Li was cited as disclosing an improved adaptive filter specifically for noise reduction in wireless mobile devices, which supplies the missing element.
    • Motivation to Combine: A POSITA would combine Li's detailed adaptive filter with Chen's system because Chen describes a high-level noise reduction goal but lacks specific implementation details. Li provides a known technique (adaptive filtering) to achieve Chen's stated purpose, making the combination a common-sense approach to implementing Chen's system to achieve the predictable result of improved noise reduction.
    • Expectation of Success: Petitioner asserted a high expectation of success, as the combination involves the simple substitution of one known element (Chen's generic noise reducer) for a more specific, known, and improved element (Li's adaptive filter) within a highly similar system.

Ground 2: Obviousness over Strömmer - Claims 1-3, 9, and 13-15 are obvious over Strömmer.

  • Prior Art Relied Upon: Strömmer (Application # 2013/0013303).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Strömmer, as a single reference, discloses all limitations of the challenged claims. Strömmer describes a system for audio signal processing in a mobile device that includes a beamformer containing both a Voice Activity Detector (VAD) and a DOA estimation unit. Petitioner asserted these components meet the limitations of the claimed "speech segment determiner" and "voice direction detector," respectively. Furthermore, Strömmer was alleged to disclose two adaptive filters that perform noise reduction: the adaptive beamformer itself, which uses DOA information to form a directional beam, and a subsequent, separate noise reduction stage that further refines the signal.
    • Key Aspects: The core of this argument is that Strömmer's integrated system inherently performs all the claimed functions. Petitioner emphasized that Strömmer's beamformer is adaptive by definition because it uses DOA information to filter signals, and the subsequent noise reduction stage is also adaptive because its attenuation level changes based on the current DOA information, thus meeting the "adaptive filter" limitation twice over.

Ground 3: Obviousness over Florencio - Claims 1-3, 9, and 13-15 are obvious over Florencio.

  • Prior Art Relied Upon: Florencio (Application # 2008/0240463).

  • Core Argument for this Ground:

    • Prior Art Mapping: Similar to the Strömmer ground, Petitioner argued that Florencio alone renders the claims obvious. Florencio describes a noise reduction system that uses a VAD to classify audio frames as speech or noise, satisfying the "speech segment determiner" limitation. When speech is detected, a Sound Source Localization (SSL) unit determines the signal's DOA, meeting the "voice direction detector" limitation. This VAD and DOA information is then fed to an "enhanced" adaptive beamformer, which performs noise reduction and thus constitutes the claimed "adaptive filter."
    • Key Aspects: Petitioner contended that Florencio’s disclosure explicitly details the functional dependency claimed in the ’259 patent: the VAD's output (speech presence) directly triggers the SSL module, which in turn provides location data to the adaptive beamformer. This direct, disclosed workflow in Florencio was argued to map precisely onto the structure of the challenged claims.

  • Additional Grounds: Petitioner asserted numerous additional obviousness challenges that build upon these core references. These grounds introduce further prior art to meet limitations in dependent claims. For example, Visser (Patent 8,897,455) was combined with the primary references to teach bypassing noise reduction when a sound source is broadside (claims 4, 7, 12, 16). Kim (a 2010 dissertation) and Kleffner (WO 2009/151578) were added to teach using a higher sampling frequency for DOA detection and a lower one for adaptive filtering (claims 10-11). Brandstein (a 2001 textbook) was combined to teach specific magnitude-based DOA techniques (claims 5-6, 17-18).

4. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-20 of Patent 9,031,259 as unpatentable under 35 U.S.C. §103.