Audio Echo Cancellation Using Beam Forming and Spectral Subtraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio systems, particularly in hands-free telecommunication devices, face challenges in effectively canceling both linear and non-linear echo components, leading to suboptimal audio quality due to the limitations of conventional echo cancellation methods which assume a linear echo path.

Innovation Solution

A device and method that separate linear and non-linear echo components using a pre-processing unit with filter units and a beam forming unit to estimate echo signals, allowing for independent control of echo cancellation through spectral subtraction, combining linear echo cancellation with beam forming to address non-linear echoes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional linear echo cancellation methods are used, then the system is simple to implement, but non-linear echo components cannot be effectively canceled leading to suboptimal audio quality

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

Solution Approach 1:

The echo cancellation system is segmented into multiple independent components: a pre-processing unit that estimates both linear and non-linear echo signals using separate filter units, and a post-processing unit that independently controls spectral subtraction for each echo type. This segmentation allows the system to address non-linear echoes without requiring complete redesign of the entire echo cancellation architecture.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges linear echo cancellation (using adaptive filters) with non-linear echo cancellation (using beam forming and spectral subtraction) into a unified system. The pre-processing unit combines outputs from both linear and non-linear echo estimation paths, allowing simultaneous cancellation of both echo types while maintaining independent control over each cancellation mechanism.

Inventive Principle:
Principle #5Merging (Combining)

2Reliability

If high over-subtraction is used to cancel echoes, then echo suppression is improved, but speech distortion increases

Engineering Contradiction:
Improveecho suppressionVSAvoidspeech quality
Core Design Contradiction:
ReliabilityVSManufacturing precision

Solution Approach 1:

The system applies different spectral subtraction strategies tailored to specific echo types: linear echo components are handled with standard spectral subtraction, while non-linear echo components use adapted spectral subtraction with modified parameters. This local quality approach ensures optimal echo suppression for each echo type without unnecessarily distorting speech components.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent dynamically adjusts spectral subtraction parameters based on the estimated echo signals. The post-processing unit modifies subtraction levels and frequency-dependent gains according to the characteristics of detected non-linear echoes, allowing aggressive echo suppression where needed while preserving speech quality in regions where echoes are less prominent.

Inventive Principle:
Principle #35Parameter changes

3Use of energy by moving object

If linear echo cancellation only is implemented, then the system remains computationally efficient, but audio quality deteriorates due to residual non-linear echoes

Engineering Contradiction:
Improvecomputational efficiencyVSAvoidaudio quality
Core Design Contradiction:
Use of energy by moving objectVSReliability

Solution Approach 1:

The pre-processing unit performs preliminary estimation of non-linear echo signals using beam forming techniques before the main echo cancellation process. By pre-identifying and separating non-linear echo components from the mixed signal, the system reduces the computational burden on subsequent processing stages and enables more efficient spectral subtraction operations.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces intermediate echo signal estimates as mediator variables between the raw microphone input and the final echo-cancelled output. The pre-processing unit generates intermediate representations of linear and non-linear echoes that facilitate more efficient processing in the post-processing unit, reducing overall computational complexity while improving audio quality.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP2171714B1A device for and a method of processing audio signals
Publication Date: 2012.08.15 KONINKLIJKE PHILIPS NV
  • EP2171714B1 patent drawingFigure 1~2
  • EP2171714B1 patent drawingFigure 3
  • EP2171714B1 patent drawingFigure 4

AI summary

A method is provided which is suitable to cope with non-linear echo paths during acoustic echo cancellation in speakerphones. Non-linear paths occur particularly in hands-free operation of, e.g., a mobile phone, due to driving the amplifier and loudspeaker in the non- linear range. The idea is to combine the commonly known one microphone approach of linear acoustic echo cancellation using an adaptive filter and a post-processor together with a multiple microphone approach using beam forming which separately removes the non- linear part of the echo.