Audio Pre-Conditioning for Machine Perception Under Echo and Noise

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio processing systems optimized for human perception often degrade the performance of machine perception systems due to differing audio processing parameters, leading to issues like echo cancellation and noise suppression that are too aggressive for machines.

Innovation Solution

Implementing pre-conditioning techniques that allow more echo leakage and background noise, and adding masking noise, tailored for machine perception, to improve the performance of automatic speech recognition systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If aggressive echo cancellation is applied to improve human perception, then human listening quality is improved, but machine perception performance deteriorates due to parameter mismatch

Engineering Contradiction:
Improvehuman listening qualityVSAvoidmachine perception performance
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The system segments the audio processing into two distinct processing paths: one optimized for human perception (with aggressive echo cancellation) and another for machine perception (with machine-optimized parameters). This allows each path to operate with parameters tailored to its specific requirements without compromising the other.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different quality standards and processing parameters are applied locally to different output streams. The human-perception stream receives heavily processed audio with aggressive echo cancellation, while the machine-perception stream receives audio processed with parameters optimized for machine recognition, allowing each to achieve its optimal performance.

Inventive Principle:
Principle #3Local quality

2Ease of operation

If noise suppression is intensified to improve human perception, then human listening experience is improved, but machine perception accuracy decreases due to loss of useful signal information

Engineering Contradiction:
Improvehuman listening experienceVSAvoidmachine perception accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The audio processing system is divided into separate processing chains for human and machine consumption. The human-perception path applies aggressive noise suppression to improve listening experience, while the machine-perception path uses milder processing to preserve signal information needed for accurate machine recognition.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different noise suppression intensities are applied locally to different output streams. The human-output stream receives heavily noise-suppressed audio for improved listening quality, while the machine-output stream receives audio with noise suppression tuned to maintain machine perception accuracy.

Inventive Principle:
Principle #3Local quality

3Ease of operation

If audio is over-processed for human perception, then human perception quality is improved, but the audio becomes less understandable to machines due to parameter optimization mismatch

Engineering Contradiction:
Improvehuman perception qualityVSAvoidmachine understandability
Core Design Contradiction:
Ease of operationVSLoss of information

Solution Approach 1:

The system creates separate processing pipelines that segment the audio flow into human-consumable and machine-consumable streams. Each pipeline applies processing parameters optimized for its target audience, preventing the loss of machine-understandability in the human-optimized stream and vice versa.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different processing quality levels are applied locally to different output destinations. The human-perception stream receives over-processed audio with parameters optimized for human ears, while the machine-perception stream receives audio processed with parameters that maintain machine understandability, ensuring each receives the appropriate quality level.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4708289A1Pre-conditioning audio for machine perception
Publication Date: 2026.03.11 DOLBY LABORATORIES LICENSING CORP
  • EP4708289A1 patent drawingFigure 1~2
  • EP4708289A1 patent drawingFigure 3
  • EP4708289A1 patent drawingFigure 4~5

AI summary

An apparatus and method of pre-conditioning audio for machine perception. Machine perception differs from human perception, and different processing parameters are used for machine perception applications (e.g., speech to text processing) as compared to those used for human perception applications (e.g., voice communications). These different parameters may result in pre-conditioned audio that is worsened for human perception yet improved for machine perception.