Coherence-Based Noise Estimation Control for Audio Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Communication systems face challenges in managing feedback loops and accurately estimating background noise, particularly when far side noise dominates, leading to distorted signals and growing noise levels, especially in systems that transmit music content.
Innovation Solution
The implementation of a feedback detector to identify signal feedback situations and limit the growth of the background noise estimate by calculating a coherence value between input and reference audio signals, allowing for adjustments to the stored noise estimate without relying on voice activity detectors, thereby preventing noise feedback loops.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If gating is applied to suppress far side noise feedback, then noise feedback is reduced, but the audio signal is distorted and sound quality deteriorates
Solution Approach 1:
The patent replaces the mechanical gating approach (binary on/off noise suppression) with a coherence-based adaptive filtering system. The system calculates coherence between reference and error signals to dynamically adjust noise suppression gain, substituting crude mechanical gating with sophisticated signal processing that preserves audio quality while suppressing feedback.
Solution Approach 2:
The system dynamically changes the noise suppression gain parameter based on coherence calculations. Instead of applying fixed gating, the gain is adjusted according to the coherence value between reference and error signals, allowing adaptive control that maintains signal integrity while suppressing feedback when coherence indicates feedback conditions.
2Reliability
If a voice activity detector is used to control feedback, then feedback loop is managed for voice signals, but the detector fails to accurately recognize non-voice signals and background noise estimation deteriorates
Solution Approach 1:
The coherence-based feedback detection mechanism serves multiple functions: it detects feedback for voice signals, detects feedback for non-voice signals, and provides accurate background noise estimation. This universal approach replaces the voice-activity-detector's limited voice-only detection with a method that works across all signal types.
Solution Approach 2:
The patent substitutes the voice activity detector's voice-based detection mechanism with a coherence-based detection system that analyzes the relationship between reference and error signals. This replacement enables accurate feedback detection and noise estimation for all signal types, not just voice.
3Object-affected harmful factors
If no voice activity detector is used, then feedback loops may occur causing continually growing background noise levels, but using one introduces inaccurate signal recognition
Solution Approach 1:
The system uses feedback from the coherence calculation between reference and error signals to control noise suppression. This feedback mechanism continuously monitors the system state and adjusts noise suppression accordingly, preventing background noise growth without relying on voice activity detection.
Solution Approach 2:
The patent replaces the voice activity detector's inaccurate recognition mechanism with a coherence-based system that accurately measures the relationship between reference and error signals. This substitution provides reliable feedback control and accurate background noise estimation without the recognition errors of voice activity detection.
Data Source
AI summary
A noise estimation control system may limit increases of a stored background noise estimate in response to a detected noise feedback situation. The system receives an input audio signal detected within a space, and a reference audio signal that is transmitted by a speaker as an aural signal into the space. A signal processor processes the input audio signal and the reference audio signal to determine a coherence value based on an amount of the aural signal that is included in the input audio signal. The signal processor also calculates an amount to adjust the stored background noise estimate based on the coherence value and a determined background noise level of the input audio signal.