Acoustic Echo Cancellation via Segmented Linear and Non-Linear Parameter Estimation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing speech communication devices face significant signal quality degradation due to interactions between linear and non-linear acoustic echo cancellation algorithms, especially in reverberant environments, where undesired signal components frequently occur, leading to mis-estimation of parameters and poor performance.

Innovation Solution

A method that generates power spectral density estimates for both linear and non-linear reverberant components, updates parameters to reduce undesired signal components through residual echo suppression, and uses a power ratio between these components as control information to adapt the acoustic echo cancellation filter, allowing concurrent identification of linear and non-linear parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional acoustic echo cancellation algorithms are used in reverberant environments, then basic echo suppression is achieved, but signal quality degrades due to interactions between linear and non-linear algorithms

Engineering Contradiction:

Improveecho suppression performanceVSAvoidsignal quality degradation

Core Design Contradiction:

ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent segments the acoustic echo cancellation process into distinct linear and non-linear components. By separating these algorithms and processing them independently with separate parameter estimation paths, the harmful interactions between linear and non-linear algorithms are eliminated, resolving the signal quality degradation while maintaining reliable echo suppression.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary mechanism in the form of separate parameter estimation paths for linear and non-linear components. This intermediary structure prevents direct interaction between the two algorithm types, allowing each to operate optimally without mutual interference, thus improving signal quality while maintaining echo suppression reliability.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If linear and non-linear algorithms are run concurrently, then comprehensive echo cancellation is achieved, but parameter mis-estimation occurs due to frequent undesired signal components

Engineering Contradiction:

Improveparameter estimation accuracyVSAvoidconcurrent algorithm operation

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The patent segments parameter estimation into separate paths for linear and non-linear algorithms. This segmentation allows each algorithm type to estimate its parameters independently without contamination from undesired signal components generated by the other algorithm type, eliminating mis-estimation while preserving concurrent operation capabilities.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by treating linear and non-linear parameter estimation differently, with each having its own optimized estimation path tailored to its specific characteristics. This localized approach ensures accurate parameter estimation for each component while maintaining the overall system's adaptability to different acoustic environments.

Inventive Principle:

Principle #3Local quality

3Object-affected harmful factors

If traditional residual echo suppression is applied, then some undesired signal components are reduced, but computational power requirements increase

Engineering Contradiction:

Improveundesired signal componentsVSAvoidcomputational power consumption

Core Design Contradiction:

Object-affected harmful factorsVSUse of energy by moving object

Solution Approach 1:

The patent segments residual echo suppression into linear and non-linear components, allowing each to be processed with computationally efficient methods appropriate to its characteristics. This segmented approach reduces the overall computational burden compared to traditional unified suppression methods while effectively reducing undesired signal components from both algorithm types.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS10481831B2System and method for combined non-linear and late echo suppression

Publication Date: 2019.11.19 CERENCE OPERATING CO

AI summary

A method, computer program product, and computer system for receiving, by a computing device, an input signal. A first power spectral density estimate may be generated for a linear reverberant component associated with the input signal. A second power spectral density estimate may be generated for a non-linear reverberant component associated with the input signal. A power spectral density estimate may be generated by combining the first power spectral density estimate for the linear reverberant component and the second power spectral density estimate for the non-linear reverberant component. One or more parameters for at least one of the linear reverberant component and the non-linear reverberant component may be updated. One or more undesired signal components in an output signal resulting from the input signal may be reduced via residual echo suppression based upon, at least in part, updating the one or more parameters.