Nonlinear Echo Masking for Residual Echo Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing acoustic echo cancellation systems fail to effectively suppress residual echo and far end single talk leakage, leading to undesirable attenuation of desired sound and poor communication quality in conferencing environments.
Innovation Solution
A non-linear processor is used to generate a continuous mask value as a gain, based on the adaptation state of an adaptive filter, coherence between microphone and filtered remote audio signals, and a comparison with a threshold, and adjusts the threshold when a loudspeaker is active to suppress residual echo and far end single talk leakage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a traditional adaptive filter is used for acoustic echo cancellation, then linear echo can be removed, but residual echo and far end single talk leakage cannot be effectively suppressed
Solution Approach 1:
The patent introduces a non-linear processor as an intermediary component between the adaptive filter and the final output. This non-linear processor generates a mask value based on coherence between microphone signal and filtered remote audio signal, and adapts the threshold dynamically. The mask value acts as a mediator that selectively suppresses residual echo and far end single talk leakage while preserving desired sound, thereby resolving the contradiction between linear echo removal and residual echo suppression.
Solution Approach 2:
The patent dynamically changes the threshold parameter of the non-linear processor based on the adaptation state of the adaptive filter and the coherence between signals. By adapting the threshold parameter in real-time, the system can effectively suppress residual echo and far end single talk leakage when conditions are favorable, while avoiding unnecessary attenuation when the adaptation state is not optimal, thus improving overall echo cancellation reliability.
2Object-generated harmful factors
If the non-linear processor suppresses residual echo aggressively, then echo cancellation improves, but desired sound may be unintentionally attenuated
Solution Approach 1:
The patent employs feedback mechanisms where the non-linear processor continuously monitors the adaptation state of the adaptive filter and the coherence between the microphone signal and the filtered remote audio signal. Based on this feedback, the processor dynamically adjusts the mask value and threshold to achieve optimal suppression of residual echo while preserving desired sound. This feedback-driven adaptation ensures that suppression actions are taken only when appropriate, maintaining communication quality.
Solution Approach 2:
The patent makes the non-linear processor dynamic by continuously adapting its threshold parameter based on the current state of the acoustic environment and the adaptation progress of the adaptive filter. Rather than using a fixed threshold, the system dynamically adjusts the threshold to match the current conditions, enabling effective residual echo suppression while avoiding attenuation of desired sound during transitions or when adaptation is incomplete.
3Reliability
If the adaptive filter is continuously adapted, then echo cancellation performance improves, but convergence time increases
Solution Approach 1:
The patent performs preliminary actions by continuously monitoring the coherence between the microphone signal and the filtered remote audio signal during the adaptation process. This preliminary coherence assessment allows the system to determine when the adaptive filter has sufficiently converged and when it is safe to activate stronger non-linear suppression, thereby reducing the overall convergence time while maintaining final performance.
Solution Approach 2:
The patent dynamically adjusts the threshold parameter of the non-linear processor based on the adaptation state of the adaptive filter. During early adaptation stages when convergence is not yet achieved, the threshold is set higher to avoid premature suppression. As adaptation progresses and convergence is detected through coherence monitoring, the threshold is lowered to enable effective residual echo suppression. This dynamic threshold adaptation reduces convergence time by enabling earlier and more effective suppression without sacrificing final performance.
Data Source
AI summary
Acoustic echo cancellation systems and methods are provided that can generate a continuous mask value that can be used as a gain of a non-linear processor. Communication between a loudspeaker and the non-linear processor can be utilized to adjust the threshold of the non-linear processor when the loudspeaker is active to assist in suppressing far end single talk leakage. The systems and methods can improve the removal of residual echo and therefore enhance the overall performance of the acoustic echo cancellation system.