Echo Cancellers Using Frequency-Domain Step-Size Adaptation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing acoustic echo cancellers face challenges with unstable behavior and slow convergence due to non-cancellable terms and misalignment issues, leading to incomplete echo reduction and distortion of near-end signals, especially in multi-channel and multi-segment systems.

Innovation Solution

A computer-implemented method for updating frequency-domain filter coefficients in an echo canceller using an optimal step size calculation based on canceller output, error, look-backward, and look-forward errors, which allows for immediate computation without delay and improves convergence and echo reduction efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional echo cancellers are used to reduce acoustic echo in multi-channel systems, then echo reduction is achieved, but the convergence is slow and behavior is unstable due to non-cancellable terms and misalignment issues

Engineering Contradiction:
Improveconvergence stabilityVSAvoidconvergence time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the echo cancellation problem into multiple independent channels and time blocks. By processing each channel separately and using block-based frequency domain processing, the system achieves more stable convergence while reducing computational complexity. The segmentation allows parallel processing that speeds up convergence without sacrificing stability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transforms the echo cancellation from time domain to frequency domain processing, changing the parameter domain to enable more efficient computation. By using frequency domain block processing with proper windowing and overlap-add techniques, the system achieves faster convergence while maintaining stability through controlled parameter updates.

Inventive Principle:
Principle #35Parameter changes

2Object-affected harmful factors

If multi-channel echo cancellation is implemented to handle complex acoustic environments, then echo reduction effectiveness is improved, but computational complexity and system requirements increase

Engineering Contradiction:
Improveecho reduction effectivenessVSAvoidsystem complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent divides the multi-channel echo cancellation into independent channel processing units, each handling a specific audio channel. This segmentation allows the system to effectively handle complex multi-channel acoustic environments while keeping each processing unit relatively simple, reducing overall system complexity through modular design.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent replaces complex time-domain convolution operations with frequency-domain multiplication operations. This substitution dramatically reduces computational complexity from O(N²) to O(N log N) per channel, enabling multi-channel echo cancellation to be practically implemented without excessive computational resources.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If frequency domain processing is used to speed up echo cancellation computation, then processing efficiency is improved, but implementation complexity and memory requirements increase

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidimplementation complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent uses block-based periodic processing in the frequency domain, where audio data is divided into fixed-size blocks processed sequentially with overlap-add. This periodic approach maintains processing efficiency while reducing memory requirements compared to full-buffer frequency domain processing, and simplifies implementation through regular, predictable operation patterns.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The patent performs preliminary windowing and block segmentation on time-domain data before transforming to frequency domain. This preliminary preparation simplifies the frequency domain processing by ensuring proper data formatting and reducing computational overhead during the actual echo cancellation operation, balancing efficiency gains with implementation simplicity.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3692703B9Echo canceller and method therefor
Publication Date: 2021.11.17 PROACTIVAUDIO GMBH
  • EP3692703B9 patent drawingFigure 1
  • EP3692703B9 patent drawingFigure 2
  • EP3692703B9 patent drawingFigure 3

AI summary

The invention relates to a computer-implemented method for updating at least one frequency-domain filter coefficient W i,j (k ) of an echo canceller having at least one channel and at least one segment per channel, the filter coefficients of the echo canceller being updateable in the frequency domain at a time block m comprising: determining a canceller output E m (k) over the mth time block; determining a canceller error ε(k) over the m— lth time block; determining a look-backward error (I); determining a look-forward error (II) as (III) wherein (IV) is based on (V); determining an optimal update step-size μ i,j,m (k) from said canceller output E m (k) over the mth time block, from said canceller error ε(k) over the m— lth time block, from said look-backward error (I), and from said look-forward error (II); and updating said at least one filter coefficient W i,j (k )by using said optimal update step-size μ i,j,m (k). The invention further pertains to an echo canceller (50) implementing said method.