Multichannel Acoustic Echo Cancellation Using Basis-Function Pre-Filters

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing multichannel nonlinear acoustic echo cancellation (NLAEC) systems are computationally demanding and unsuitable for miniaturized devices, such as smart speaker devices, due to the estimation of simultaneously active nonlinearities and the need for multiple microphones, which degrades the quality of near-end speaker's speech.

Innovation Solution

An acoustic processing device that applies pre-processing filters based on a linear combination of pre-defined basis functions to each loudspeaker signal, estimating echo signals for each microphone signal, and uses a scalable approach with adjustable weighting coefficients to minimize computational cost, allowing for effective echo cancellation without degrading speech quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional multichannel nonlinear acoustic echo cancellation methods are used, then echo cancellation capability is improved, but computational complexity increases making the system unsuitable for miniaturized devices

Engineering Contradiction:
Improveecho cancellation capabilityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the nonlinear echo cancellation problem into multiple independent linear echo cancellation subproblems by using piecewise linear approximation. Each segment uses a linear model with reduced parameters, transforming one complex nonlinear estimation problem into multiple simpler linear problems that can be solved independently with lower computational cost.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameter representation by using piecewise linear approximation with breakpoints. Instead of directly estimating complex nonlinear parameters, the system estimates linear parameters for each segment and combines them, reducing the overall computational burden while maintaining cancellation effectiveness.

Inventive Principle:
Principle #35Parameter changes

2Reliability

If estimation is carried out for multiple microphones to cancel nonlinear echo, then echo cancellation performance is improved, but speech quality of near-end speaker deteriorates due to computational demands

Engineering Contradiction:
Improveecho cancellation performanceVSAvoidspeech quality
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

The patent applies segmentation to divide the echo cancellation across multiple microphones into independent linear subproblems. This allows parallel processing with reduced computational interference between microphone channels, preserving speech quality while achieving multi-microphone echo cancellation performance.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces piecewise linear approximation as an intermediary model between the original nonlinear echo path and the final echo cancellation. This intermediary linear model reduces computational complexity and interference in multi-microphone processing, thereby preserving near-end speaker speech quality while maintaining cancellation effectiveness.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4205108B1Acoustic processing device for multichannel nonlinear acoustic echo cancellation
Publication Date: 2026.01.28 HUAWEI TECH CO LTD
  • EP4205108B1 patent drawingFigure 1
  • EP4205108B1 patent drawingFigure 2a
  • EP4205108B1 patent drawingFigure 2b

AI summary

An acoustic processing device (110) for multichannel nonlinear acoustic echo cancellation is disclosed. The acoustic processing device (110) comprises a processing circuitry (120) configured to apply to each of a plurality of loudspeaker signals a respective pre- processing filter for filtering each loudspeaker signal in order to obtain a respective pre-processed loudspeaker signal, wherein each pre-processing filter is based on a linear combination of a plurality of pre-defined basis functions, wherein each of the pre-defined basis functions is weighted by an adjustable pre-processing weighting coefficient. The processing circuitry (120) is further adapted to enable echo reduction by determining for each of a plurality of microphone signals an estimated echo signal on the basis of the plurality of pre-processed loudspeaker signals and determining a respective echo reduced microphone signal based on the respective microphone signal and the estimated echo signal.