Dual-Filter Echo Cancellation for Double-Talk and Echo Path Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Echo cancellers face challenges in efficiently re-adapting to changes in the echo path while maintaining effective echo attenuation, especially in noisy conditions and double-talk scenarios, due to high computational load and difficulties in determining filter performance differences.
Innovation Solution
An echo canceller employing an adaptive main filter and a faster adaptive shadow filter, where the shadow filter's adaptation speed is used to update the system noise estimate of the main filter, allowing for robust discrimination between double talk and echo path changes, with a dynamic Q factor adjusting the adaptation speed based on residual signal differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a Kalman filter is used for implementing the linear filter adaptation, then the echo cancellation performance is improved, but the computational load becomes prohibitively large for real-time implementation
Solution Approach 1:
The patent uses a shadow filter that is discarded after a short adaptation period once the main filter has learned from it. The shadow filter serves as a temporary, low-cost mechanism to quickly capture echo path changes without the computational burden of continuously running a full Kalman filter. This allows the system to benefit from fast adaptation capabilities while maintaining acceptable computational load for real-time operation.
Solution Approach 2:
The patent segments the adaptation process into two distinct phases: a fast shadow filter adaptation phase and a slower main filter update phase. This segmentation allows the system to separate the computationally intensive tasks, using the shadow filter to quickly detect changes and then updating the main filter only when necessary, thereby reducing overall computational load while maintaining echo cancellation performance.
2Adaptability or versatility
If the adaptation speed of the main filter is increased to quickly track echo path changes, then the re-adaptation capability is improved, but the echo attenuation performance deteriorates in noisy conditions
Solution Approach 1:
The shadow filter acts as an intermediary between the acoustic environment and the main filter. It quickly adapts to echo path changes and serves as a mediator that provides update information to the main filter without directly controlling its adaptation. This intermediary mechanism allows the main filter to benefit from fast change detection while maintaining its noise-resistant properties through controlled adaptation speed.
Solution Approach 2:
The shadow filter performs preliminary adaptation work by quickly tracking echo path changes before the main filter needs to update. This preliminary action captures the essential change information, allowing the main filter to make informed updates at its own slower, more stable pace, thereby maintaining both fast re-adaptation capability and reliable echo attenuation.
3Difficulty of detecting and measuring
If a dual-filter structure is used to discriminate double talk from echo path changes, then the discrimination capability is improved, but the device complexity increases
Solution Approach 1:
The patent implements a dynamic filter structure where the shadow filter is actively adapted during double-talk detection periods and then discarded after contributing its adaptation information. This dynamic approach allows the system to have complex dual-filter functionality when needed for discrimination, while simplifying to a single active filter during normal operation, thereby reducing overall structural complexity while maintaining discrimination capability.
Solution Approach 2:
The shadow filter is temporarily activated when double-talk discrimination is needed, performs its function, and then is discarded. This discarding and recovering mechanism allows the system to have dual-filter discrimination capability on-demand rather than maintaining permanent dual-filter complexity, reducing overall device complexity while preserving the ability to discriminate between double-talk and echo path changes when required.
Data Source
AI summary
The present invention relates to an echo canceller (300) for estimating a model of an echo signal and a method thereof. The echo canceller comprises at least an adaptive main filter (301) for modeling the echo signal and an adaptive shadow filter (302) for modeling the echo signal. The adaptive main filter is an adaptive filter for which the adaptation speed is proportional to a system noise estimate and the adaptation speed of the shadow filter (302) is faster than the adaptation speed of the adaptive main filter (301). The echo canceller (300) comprises a processor (303) for determining whether the adaptive shadow filter models the echo signal better than the adaptive main filter and an updater (304) for updating the system noise estimate of the adaptive main filter (301) if the adaptive shadow filter models (302) the echo signal better than the adaptive main filter (301).