Method and apparatus for cross-talk resistant adaptive noise canceller

• US 8,340,278 B2
• Filed: 10/07/2010
• Issued: 12/25/2012
• Est. Priority Date: 11/20/2009
• Status: Active Grant

• Current Assignee: Texas Instruments, Inc.
• Original Assignee: Texas Instruments, Inc.
• Inventors: Gowreesunker, Baboo Vikrhamsingh, Unno, Takahiro, Ikram, Muhammad Zubair
• Primary Examiner(s): Anwah, Olisa
• Application Number:
  US12/900,143
• Publication Number:
  US 20110123019A1
• Time in Patent Office: 810 Days
• Field of Search: 379/406.09, 379/406.03, 379/406.01, 379/406.02, 379/406.08
• US Class Current: 379/406.09
• CPC Class Codes: H04B 3/23 using a replica of transmitted signal in the time domain, e.g. echo cancellers

Abstract

A method and apparatus for cross-talk resistant adaptive noise cancellation. The method includes retrieving, via the processor, a primary signal and a reference signal, filtering the primary signal utilizing a filter H₁₂(z) and estimating a cross-talk, filtering the reference signal utilizing a filter H₂₁(z) and estimating the noise leakage of the reference signal, determining the difference between the noise leakage estimate from the primary signal and estimating a first post-filtering, determining the difference between the cross-talk estimate and the reference signal and estimating a second post-filtering, determining if the voice activity is detected in the primary signal, adapting filters H₁₂ and H₂₁ by de-correlation if the voice activity is detected, and adapting filter H₁₂ by NLMS if the voice activity is not detected, limiting maximum filter change during different conditions for protecting filters H₁₂ and H₂₁ from diverging, maintaining filter stability by tracking absolute sum of the coefficients of the convolution of H₁₂ and H₂₁, obtaining a primary output and a reference output utilizing the first post-filtering and the second post-filtering, and utilizing the primary output and the reference output for cross-talk resistant adaptive noise cancellation.

Claims

1. A method of a processor for cross-talk resistant adaptive noise cancelling, comprising:

retrieving, via the processor, a primary signal and a reference signal;

filtering the primary signal utilizing a filter H₁₂(z) and estimating a cross-talk;

filtering the reference signal utilizing a filter H₂₁(z) and estimating the noise leakage of the reference signal;

determining the difference between the noise leakage estimate from the primary signal and estimating a first post-filtering;

determining the difference between the cross-talk estimate and the reference signal and estimating a second post-filtering;

determining if the voice activity is detected in the primary signal;

adapting filters H₁₂ and H₂₁ by de-correlation if the voice activity is detected, and adapting filter H₁₂ by NLMS if the voice activity is not detected;

limiting maximum filter change during different conditions for protecting filters H₁₂ and H₂₁ from diverging;

maintaining filter stability by tracking absolute sum of the coefficients of the convolution of H₁₂ and H₂₁;

obtaining a primary output and a reference output utilizing the first post-filtering and the second post-filtering; and

utilizing the primary output and the reference output for cross-talk resistant adaptive noise cancellation.

2. An apparatus for cross-talk resistant adaptive noise cancelling, comprising:

means for retrieving a primary signal and a reference signal;

means for filtering the primary signal utilizing a filter H₁₂(z) and estimating a cross-talk;

means for filtering the reference signal utilizing a filter H₂₁(z) and means for estimating the noise leakage of the reference signal;

means for determining the difference between the noise leakage and the estimate from the primary signal and means for estimating a first post-filtering;

means for determining the difference between the cross-talk estimate and the reference signal and means for estimating a second post-filtering;

means for determining if the voice activity is detected in the primary signal;

means for adapting filters H₁₂ and H₂₁ by de-correlation if voice activity is detected, and means for adapting filter H₁₂ by NLMS if voice activity is not detected;

means for limiting maximum filter power change for protecting filters H₁₂ and H₂₁ from diverging;

means for maintaining stability of the system by limiting the absolute sum of the coefficients of the convolution of H₁₂ and H₂₁;

means for obtaining a primary output and a reference output utilizing the first post-filtering and the second post-filtering; and

means for utilizing the primary output and the reference output for cross-talk resistant adaptive noise cancellation.

3. A non-transitory computer readable medium comprising computer instructions, when executed perform a method for cross-talk resistant adaptive noise cancelling, the method comprising:

retrieving, via the processor, a primary signal and a reference signal;

filtering the primary signal utilizing a filter H₁₂(z) and estimating a cross-talk;

filtering the reference signal utilizing a filter H₂₁(z) and estimating the noise leakage of the reference signal;

determining the difference between the noise leakage estimate from the primary signal and estimating a first post-filtering;

determining the difference between the cross-talk estimate and the reference signal and estimating a second post-filtering;

determining if the voice activity is detected in the primary signal;

adapting filters H₁₂ and H₂₁ by de-correlation if the voice activity is detected, and adapting filter H₁₂ by NLMS if the voice activity is not detected;

limiting maximum filter change during different conditions for protecting filters H₁₂ and H₂₁ from diverging;

maintaining filter stability by tracking absolute sum of the coefficients of the convolution of H₁₂ and H₂₁;

obtaining a primary output and a reference output utilizing the first post-filtering and the second post-filtering; and

utilizing the primary output and the reference output for cross-talk resistant adaptive noise cancellation.