Echo Suppression Filter Control for Stationary Noise Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional acoustic echo suppression systems fail to effectively handle stationary noise in microphone signals, leading to systematic estimate deviations and degraded audio quality, particularly in scenarios with low signal-to-noise ratios.
Innovation Solution
The proposed solution involves computing control information for a suppression filter by determining energy-related values for band-pass signals from temporally successive data blocks, modifying these values based on mean values, and using the modified values to compute control information for the suppression filter, thereby separating stationary noise from actual signals and improving echo suppression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional acoustic echo suppression systems process microphone signals directly, then the processing is simple and fast, but stationary noise causes systematic estimate deviations and degraded audio quality
Solution Approach 1:
The patent segments the microphone signal into multiple frequency bands using band-pass filters. By processing each frequency band separately and computing energy-related values for each band, the system can identify and handle stationary noise more effectively in each band, improving overall echo estimation accuracy without excessively increasing complexity.
Solution Approach 2:
The patent performs preliminary processing by determining energy-related values for band-pass signals from temporally successive data blocks before computing the final echo control information. This preliminary action includes computing mean values and modifying energy-related values to account for stationary noise, which improves subsequent echo estimation accuracy.
2Productivity
If energy-related values are computed from raw microphone signals, then the computation is straightforward, but stationary noise degrades the quality of control information
Solution Approach 1:
The patent performs preliminary computation of energy-related values for band-pass signals from temporally successive data blocks. By computing mean values of these energy-related values and modifying them to account for stationary noise components, the system prepares more reliable control information in advance, which improves the reliability of subsequent echo suppression without significantly impacting computation speed.
3Device complexity
If the suppression filter processes all frequency components equally, then the processing is simple, but stationary noise cannot be effectively separated from useful signals
Solution Approach 1:
The patent segments the signal processing into multiple frequency bands using band-pass filters. By computing energy-related values and their mean values separately for each frequency band, the system can identify stationary noise characteristics in each band and effectively separate them from useful signals, improving signal separation accuracy while maintaining manageable processing complexity.
Data Source
AI summary
An embodiment of an apparatus for computing control information for a suppression filter for filtering a second audio signal to suppress an echo based on a first audio signal includes a computer having a value determiner for determining at least one energy-related value for a band-pass signal of at least two temporally successive data blocks of at least one signal of a group of signals. The computer further includes a mean value determiner for determining at least one mean value of the at least one determined energy-related value for the band-pass signal. The computer further includes a modifier for modifying the at least one energy-related value for the band-pass signal on the basis of the determined mean value for the band-pass signal. The computer further includes a control information computer for computing the control information for the suppression filter on the basis of the at least one modified energy-related value.