Hearing Aid Neural Feedback Control for Stable High Gain
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing NLMS-based feedback control systems in hearing aids suffer from biased estimation issues, leading to suboptimal gain provision and poor sound quality, especially for musicians and super users, and struggle to handle critical feedback situations quickly.
Innovation Solution
Implementing a machine learning-based feedback control system using neural networks for feedback cancellation, trained with synthetic data to minimize feedback without compromising sound quality, and enhance the hearing aid's ability to handle critical feedback scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If NLMS based feedback control systems are used, then feedback cancellation is achieved, but biased estimation problem occurs leading to suboptimal gain provision
Solution Approach 1:
The patent replaces the traditional NLMS adaptive filter algorithm with a deep neural network-based feedback cancellation system. The DNN model learns optimal feedback cancellation strategies from training data, substituting the conventional iterative adaptation mechanism with a pre-trained neural network that provides more accurate feedback path estimation without biased estimation issues.
Solution Approach 2:
The system performs preliminary training of the deep neural network offline using synthetic feedback scenarios before actual hearing aid operation. This preliminary action allows the model to learn optimal feedback cancellation strategies in advance, enabling accurate feedback estimation during real-time operation without requiring complex adaptive tuning during actual use.
2Measurement precision
If frequency shift and probe noise methods are introduced to solve biased estimation, then estimation accuracy improves, but sound quality deteriorates
Solution Approach 1:
The patent eliminates the need for frequency shift and probe noise injection methods by replacing the NLMS adaptive filter with a deep neural network. The DNN directly estimates the feedback path using learned patterns from training data, achieving accurate feedback cancellation without introducing additional noise or frequency shifts that would degrade sound quality.
Solution Approach 2:
The system uses synthetic training data that copies realistic feedback scenarios to train the neural network. This allows the model to learn from simulated examples without requiring actual probe noise or frequency shifts during operation, thereby maintaining sound quality while achieving accurate feedback estimation.
3Stability of the object's composition
If traditional feedback control systems are used, then system stability is maintained, but response speed to critical feedback situations is insufficient
Solution Approach 1:
The patent replaces the iterative adaptive filtering process with a deep neural network that provides direct feedback cancellation. The DNN's pre-trained weights enable immediate response to feedback situations without requiring multiple adaptation iterations, significantly improving response speed while maintaining system stability through the learned cancellation strategy.
Solution Approach 2:
The system performs preliminary training of the neural network offline to prepare optimal response strategies in advance. This preliminary action enables the feedback control system to respond immediately to critical situations during operation, eliminating the delay associated with real-time adaptive tuning in traditional systems.
4Reliability
If gain is reduced to avoid feedback, then feedback stability is improved, but sound quality and user satisfaction deteriorate
Solution Approach 1:
The patent replaces conventional feedback avoidance strategies with a deep neural network-based feedback cancellation system. The DNN accurately estimates and cancels feedback paths, allowing the system to maintain high gain levels for optimal sound quality while achieving feedback stability through intelligent cancellation rather than conservative gain reduction.
Data Source
AI summary
A hearing aid comprises a) at least one input transducer for providing at least one electric input signal representing said sound; b) an output transducer for providing stimuli perceivable to the user as sound; c) a feedback control system configured to minimize feedback from said output transducer to said at least one input transducer, and to at least provide a feedback corrected version of said at least one electric input signal; and d) an audio signal processor configured to apply one or more processing algorithms to said feedback corrected version of said at least one electric input signal, and to provide a processed signal in dependence thereof. The feedback control system is based on a machine learning model receiving input data at least representing said at least one electric input signal; and said processed signal; and providing said feedback corrected version of the at least one electric input signal as an output. A method of training a machine learning model is further disclosed.