Adaptive Feedback Estimation System with Correlation Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio processing devices, such as hearing aids, face challenges in minimizing the audibility of artefacts caused by acoustic feedback, especially in environments with tonal components like music, due to biased feedback path estimates from adaptive filters when the input signal is not white.
Innovation Solution
An audio processing device that incorporates a correlation detection unit to provide correlation measures between input and output signals, and a de-correlation unit that applies a frequency shift to improve feedback estimation, with a control unit managing the adaptive algorithm and de-correlation unit to minimize artefact audibility by adjusting the adaptation rate and frequency shift based on auto-correlation and cross-correlation values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If adaptive feedback cancellation is used to track feedback path changes over time, then system stability is improved, but artefacts are introduced when there is autocorrelation in the input signal
Solution Approach 1:
The system performs preliminary detection of autocorrelation in the input signal before executing the adaptive filter update. When autocorrelation is detected, the system prepares by disabling the update mechanism in advance, preventing the biased feedback path estimate and resulting artefacts from occurring
Solution Approach 2:
The system uses feedback from the autocorrelation detector to control the adaptive algorithm update process. The detector continuously monitors the input signal characteristics and provides feedback to the update controller, which adjusts the update rate or disables updates when autocorrelation is present, thereby preventing artefact generation while maintaining stability during normal operation
2Speed
If the adaptation rate of the adaptive algorithm is increased to track feedback path changes faster, then responsiveness is improved, but estimation accuracy deteriorates in the presence of autocorrelation
Solution Approach 1:
The system dynamically adjusts the adaptation rate based on real-time analysis of the input signal characteristics. The update controller modifies the adaptation speed according to the detected autocorrelation level, allowing fast tracking when signals are white and preventing accuracy degradation when autocorrelation is present
Solution Approach 2:
The system changes the adaptation rate parameter dynamically based on signal conditions. When autocorrelation is detected, the parameter is adjusted to reduce or pause updates, thereby maintaining estimation accuracy while allowing faster adaptation during periods of white noise input
3Measurement precision
If de-correlation techniques are applied to improve feedback estimation, then estimation quality is improved, but audibility of artefacts increases in tonal environments
Solution Approach 1:
The system performs preliminary detection of tonal components in the input signal before applying de-correlation techniques. When tones are detected, the system prepares by reducing or disabling de-correlation processing in advance, preventing the increased audibility of artefacts while maintaining good estimation quality during non-tonal periods
Solution Approach 2:
The system uses feedback from the tone detector to control the de-correlation unit. The detector continuously monitors for tonal components and provides feedback to adjust the de-correlation strength, allowing improved estimation quality when signals are non-tonal while preventing artefact audibility when tones are present
Data Source
Figure 1a~1c
Figure 1d
Figure 2a~2b
AI summary
Audio feedback estimation system comprising an input transducer for picking up a sound signal and an output transducer for providing an output sound, a forward path being defined there between. It comprises also a signal processing unit for processing an electric input signal, a de-correlation unit for de-correlating the output signal from the input signal, an adaptive filter comprising a variable filter part for filtering an input signal according to variable filter coefficients and an algorithm part comprising an adaptive algorithm for dynamically updating said filter coefficients,a control unit for controlling said de-correlation unit and said adaptive algorithm, and a correlation detection unit for determining the auto-correlation of a signal of the forward path and providing an AC-value and/or the cross-correlation between two different signals of the forward path and providing an XC-value.