Audio Pre-Conditioning for Machine Perception Under Echo and Noise
Find Innovative SolutionsGenerate 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
Engineering 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
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.
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.
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
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.
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.
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
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.
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.
Data Source
Figure 1~2
Figure 3
Figure 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.