Prior art — US Patent 8320575

Analysis of Relevant Prior Art for US Patent 8,320,575

The following analysis details the most relevant prior art cited during the prosecution of US Patent 8,320,575. For a reference to anticipate a claim under 35 U.S.C. § 102, it must disclose, either expressly or inherently, every element of that claim. The analysis below focuses on potential anticipation of the independent claims, particularly claim 1 and claim 10, which form the broadest basis of the invention. The filing date for US 8,320,575 is September 30, 2008, with a priority date of October 1, 2007. Any prior art must predate this priority date.

1. US Patent 5,272,695 A: "Subband echo canceller with adjustable coefficients using a series of step sizes"

Full Citation: US Patent 5,272,695 A, filed by Nippon Telegraph And Telephone Corporation.
Publication/Filing Date: Filed September 14, 1990; Published December 21, 1993.
Brief Description: This patent describes an echo canceller that operates in the sub-band domain. An input signal is divided into multiple frequency sub-band signals by a filter bank. Adaptive filters then process these sub-band signals to cancel echo. The coefficients of these filters are updated using an algorithm that adjusts the step size for adaptation, aiming to improve convergence speed and reduce misadjustment. The processed sub-band signals are then synthesized back into a full-band signal.
Potential Anticipation of Claims:
- Claim 1: This reference appears to disclose several key elements of claim 1. It teaches dividing an audio signal into audio sub-band signals and processing a...subset of the audio sub-band signals (in this case, for echo cancellation). It also describes synthesizing the processed sub-band signals back into an enhanced audio signal. However, US 5,272,695 does not appear to explicitly teach the step of "excising a subset of the audio sub-band signals" for computational efficiency and then later "reconstructing at least a portion of the subset...that were excised." The '695 patent processes all the sub-bands generated by its filter bank. Therefore, it does not fully anticipate claim 1.
- Claim 10: This reference is highly relevant as it deals with echo cancellation using a reference signal (the far-end signal that causes the echo). It discloses dividing both the primary (microphone) and reference signals into sub-bands and using the reference sub-bands to adapt an echo compensation filter. However, for the same reason it does not anticipate claim 1, it fails to anticipate claim 10. It lacks the specific steps of excising and then reconstructing a subset of sub-band signals for either the primary or the reference signal.

2. US Patent Application Publication 2009/0119096 A1: "Partial speech reconstruction"

Full Citation: US Patent Application Publication 2009/0119096 A1, invented by Franz Gerl.
Publication/Filing Date: Filed October 29, 2007; Published May 7, 2009.
Brief Description: This application describes a method for processing a speech signal where the signal is converted into a spectral representation (frequency sub-bands). To reduce computational complexity, only a subset of these spectral components is transmitted or processed. The missing spectral components are then reconstructed at the receiving end by interpolating or extrapolating from the components that were processed. This is aimed at applications like speech coding and recognition where bandwidth or processing power is limited.
Potential Anticipation of Claims:
- Claim 1: This reference discloses the core concepts of processing a signal in sub-bands, discarding (or not processing) a portion of them, and then reconstructing the missing parts. It teaches converting a signal into spectral components (dividing), processing only a part of the spectrum (excising a subset, implicitly), and reconstructing the missing components. The final step of synthesizing them into a full-band signal is also inherent. This publication is highly relevant and could be argued to anticipate the key steps of claim 1, as the motivation (reducing computational load) and the general method (excising and reconstructing) are similar.
- Claim 10: While US 2009/0119096 A1 teaches the general process of claim 1, it does not specifically describe using this method in the context of echo cancellation with a reference signal. It does not mention dividing a reference signal into sub-bands, excising corresponding sub-bands, and using the remainder to adapt an echo compensation filter. Therefore, it does not anticipate the specific combination of elements in claim 10.

3. US Patent 6,898,235 B1: "Wideband communication intercept and direction finding device using hyperchannelization"

Full Citation: US Patent 6,898,235 B1, assigned to Argon St Incorporated.
Publication/Filing Date: Filed December 10, 1999; Published May 24, 2005.
Brief Description: This patent details a method for processing wideband signals by dividing them into a large number of narrow sub-bands (hyperchannelization). A key feature is the ability to process only a selected subset of these channels, which can be chosen based on signal activity or other criteria. This reduces the overall computational load required for signal analysis. The patent mentions that the selected channels can be processed for tasks like direction finding.
Potential Anticipation of Claims:
- Claim 1: This patent teaches dividing a signal into sub-bands and excising a subset by selecting only certain channels for further processing. The motivation is explicitly to reduce computational load. It also mentions that the processed signals can be used to create an output. However, it does not explicitly teach the step of reconstructing the excised sub-band signals to form a complete, enhanced full-band audio signal. The focus is on analyzing the selected bands rather than recreating a full-fidelity version of the original signal. Therefore, it does not fully anticipate claim 1.
- Claim 10: This patent is not directed at echo cancellation and does not disclose the use of a reference signal as described in claim 10. Thus, it does not anticipate this claim.

4. US Patent Application Publication 2008/0140396 A1: "Model-based signal enhancement system"

Full Citation: US Patent Application Publication 2008/0140396 A1, invented by Dominik Grosse-Schulte.
Publication/Filing Date: Filed October 31, 2006; Published June 12, 2008.
Brief Description: This publication describes a system for enhancing speech signals by separating them from noise. The signal is transformed into the frequency domain (sub-bands). A model of the speech signal is used to identify and preserve speech components while suppressing noise. The system can operate on selected frequency bands to improve efficiency.
Potential Anticipation of Claims:
- Claim 1: This reference teaches signal enhancement in the sub-band domain. It discloses dividing the signal and processing the sub-bands. It mentions that processing can be focused on selected bands, which implies the concept of excising others. However, similar to the other references, the explicit step of reconstructing the excised bands by interpolating or averaging from the processed bands, followed by synthesis, is not clearly taught. The focus is on noise suppression in the retained bands. Thus, it does not fully anticipate claim 1.
- Claim 10: This reference is focused on noise reduction, not echo cancellation involving a reference signal in the manner specified by claim 10. Therefore, it does not anticipate this claim.