PTAB

IPR2016-00459

Waves Audio Ltd v. Andrea Electronics Corp

Log In
Key Events
Petition
petition Intelligence

1. Case Identification

2. Patent Overview

  • Title: System, Method and Apparatus For Cancelling Noise
  • Brief Description: The ’345 patent discloses a digital signal processing system for noise cancellation in an audio signal. The system generates a frequency spectrum of the signal, creating discrete frequency bins, and then uses a threshold detector with a noise estimation process to identify the position of noise elements within each bin before subtracting them from the signal.

3. Grounds for Unpatentability

Ground 1: Anticipation of Claims 1-3, 13, and 38 by Higgins

  • Prior Art Relied Upon: Higgins (Patent 6,266,633).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Higgins discloses every limitation of the challenged claims. Higgins describes a pre-processor for noise suppression that receives a noisy audio signal and uses a Fast Fourier Transform (FFT)—the claimed “frequency spectrum generator”—to convert the signal into a frequency domain representation with multiple “frequency bins.” Higgins further teaches a noise estimation process that generates a histogram of spectral magnitudes to determine a noise threshold for each frequency. The system then performs spectral subtraction by subtracting this noise floor from the signal, which inherently detects the position of noise elements when the magnitude of a frequency bin is less than the threshold. The subtraction itself is performed by a “subtractor” module.

Ground 2: Anticipation of Claims 1-3, 12, 13, 21, 23, and 38 by Hirsch

  • Prior Art Relied Upon: Hirsch (“Noise estimation techniques for robust speech recognition,” a 1995 IEEE publication).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner asserted that Hirsch teaches all features of the challenged claims. Hirsch describes techniques to estimate noise spectra in noisy speech signals without explicit speech pause detection. It uses an FFT filter bank to divide the signal into “subbands,” equivalent to the claimed frequency bins. The core of Hirsch’s first algorithm is a noise estimation process that calculates the noise level in each subband as a weighted average of past spectral magnitude values that fall below an adaptive threshold. This comparison of the signal’s spectral component to the threshold serves as the claimed “threshold detector” that identifies noise. Hirsch further discloses that its techniques are applied as a preprocessing step for a “nonlinear spectral subtraction scheme,” thereby teaching the claimed subtractor.

Ground 3: Obviousness of Claims 1 and 38 over Higgins in view of Hirsch

  • Prior Art Relied Upon: Higgins (Patent 6,266,633) and Hirsch (a 1995 IEEE publication).

  • Core Argument for this Ground:

    • Prior Art Mapping: Petitioner argued that Higgins teaches nearly all elements of the claims, including a complete noise cancellation system based on spectral subtraction. However, should the Board find that Higgins only teaches a single noise threshold rather than a distinct threshold for each frequency bin, Hirsch explicitly remedies this alleged deficiency. Hirsch discloses calculating an adaptive noise threshold that is different for each frequency bin to achieve superior noise detection.
    • Motivation to Combine: A POSITA would combine Higgins and Hirsch because both references are directed to the same problem of suppressing noise in speech by comparing a signal to a noise threshold in the frequency domain. A POSITA would have been motivated to incorporate Hirsch’s technique of using an adaptive, per-bin threshold into the system of Higgins to improve its performance and achieve more accurate noise detection, as suggested by Hirsch.
    • Expectation of Success: A POSITA would have had a reasonable expectation of success in combining the references. Both operate on well-understood principles of spectral analysis and subtraction, and integrating an improved thresholding method (Hirsch) into a base system (Higgins) would have been a predictable modification.

  • Additional Grounds: Petitioner asserted numerous additional anticipation and obviousness challenges. These included anticipation by McAulay (a 1980 IEEE publication) and obviousness over Arslan (Patent 5,706,395). Further obviousness grounds relied on combinations of Higgins or Hirsch with references including Yang, Martin, Boll, Adams, O’Hagan, Lindemann, and Cezanne to teach limitations related to averaging, residual noise processing, magnitude estimation, smoothing, and microphone arrays.

4. Key Claim Construction Positions

  • "frequency bins": Petitioner proposed this term means "frequency domain outputs extending between two limiting frequencies." This construction was central to mapping prior art disclosures of FFT outputs or filter bank subbands to the claimed invention.
  • "threshold detector": Petitioner proposed this term means "an algorithm that, for each frequency bin, sets a threshold and compares a magnitude of the frequency bin to the predetermined threshold." This construction was critical for arguing that prior art references which compare spectral components to a calculated noise estimate meet the limitation of detecting noise.

5. Relief Requested

  • Petitioner requested institution of an inter partes review and cancellation of claims 1-3, 12-14, 17, 21, 23, 25, 38, and 47 of the ’345 patent as unpatentable.