Neural Network Gain Margin Prediction for Hearing Feedback Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing aid technologies struggle to effectively predict and reduce feedback distortions, such as howling and chirping, due to inefficient gain margin estimation and reactive feedback mitigation strategies, which can be computationally expensive and introduce audible distortions.
Innovation Solution
Implementing a deep neural network (DNN) to predict the instantaneous gain margin in hearing devices using inputs from microphone and loudspeaker signals, adaptive filter coefficients, and inertial data, allowing for proactive adjustment of feedback reduction parameters like step-size and gain to prevent feedback onset.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional gain margin estimation methods are used, then feedback can be detected, but the computational cost is high and feedback onset cannot be prevented in time
Solution Approach 1:
The neural network predicts the gain margin before feedback actually occurs, enabling proactive adjustment of feedback reduction parameters. This preliminary prediction allows the system to prevent feedback onset rather than merely detecting it after the fact, resolving the time delay issue in traditional reactive approaches.
Solution Approach 2:
The patent replaces traditional computational gain margin estimation methods with a neural network-based prediction system. This substitution enables faster, more efficient prediction of gain margin values without the high computational overhead of conventional methods, addressing both the reliability and time loss contradictions.
2Object-affected harmful factors
If reactive feedback mitigation strategies are applied, then feedback can be reduced, but audible distortions are introduced and computational resources are consumed
Solution Approach 1:
The system applies preliminary anti-action by predicting the gain margin and adjusting feedback reduction parameters before feedback distortions occur. This proactive approach prevents the harmful feedback effects from developing in the first place, rather than attempting to mitigate them after they have already been generated, thus avoiding the introduction of additional audible distortions.
Solution Approach 2:
By predicting gain margin values in advance and proactively adjusting feedback reduction parameters, the system prevents feedback onset before it can generate harmful distortions. This eliminates the need for reactive mitigation that would otherwise introduce additional audible artifacts into the audio signal.
3Reliability
If complex feedback reduction algorithms are used, then feedback can be controlled, but device complexity and computational overhead increase
Solution Approach 1:
The patent substitutes complex traditional feedback control algorithms with a trained neural network model. Once trained, the neural network provides efficient predictions with lower computational overhead during operation, maintaining feedback control effectiveness while reducing the computational burden on the hearing device.
Solution Approach 2:
The system changes the operational parameters by using neural network predictions of gain margin to dynamically adjust feedback reduction parameters. This parameter-based control approach maintains effectiveness while being computationally more efficient than complex traditional algorithms, as the neural network has already processed the complexity during the training phase.
Data Source
AI summary
A hearing device includes a microphone that produces an audio input signal and a loudspeaker that outputs an amplified audio signal into an ear canal. A signal processing path is coupled to the microphone and the loudspeaker. The signal processing path includes a deep neural network configured to predict an instantaneous gain margin of the hearing device based on a set of inputs. The set of inputs includes a first parameter of the audio input signal and a second parameter of the amplified audio signal. A feedback reduction module of the device receives the predicted instantaneous gain margin and adjusts feedback reduction parameters to reduce an onset of feedback in the hearing device