Hearing Aid Noise Reduction Using Neural SNR-to-Gain Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing aids face challenges in providing mathematically optimal noise reduction solutions that are not well received in terms of loudness perception, and the computational limitations of hearing devices make it impractical to implement large neural networks for signal-to-noise ratio (SNR) to gain conversion.
Innovation Solution
The use of machine learning techniques, specifically neural networks, to determine gain based on signal-to-noise ratio, where the weights of the neural network are trained with a plurality of training signals, allowing for efficient SNR-to-gain conversion in a computationally feasible manner, even in devices with limited capacity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If mathematically optimal noise reduction algorithms are used, then noise reduction performance is improved, but loudness perception quality deteriorates
Solution Approach 1:
The patent changes the parameters of the noise reduction algorithm by introducing a perceptual weighting function that modifies the gain application based on human loudness perception characteristics. This transforms the mathematical optimization parameters to include perceptual quality metrics, allowing the system to maintain noise reduction effectiveness while improving loudness perception by adjusting how gains are applied across different frequency bands and time frames.
2Measurement precision
If large neural networks are implemented for SNR-to-gain conversion, then noise reduction accuracy is improved, but device complexity and power consumption increase beyond practical limits
Solution Approach 1:
The patent segments the neural network into smaller, more manageable components that can be efficiently implemented in hearing aid devices. By dividing the complex SNR-to-gain conversion task into modular processing stages with localized computations, the system achieves comparable accuracy to large networks while reducing overall device complexity and power consumption to practical levels.
Solution Approach 2:
The patent applies local quality optimization by implementing computationally efficient neural network operations that focus computational resources on the most critical frequency regions and time frames. This allows accurate SNR-to-gain conversion where it matters most while using simplified computations in less critical regions, maintaining overall accuracy without requiring large-scale network implementation.
Data Source
AI summary
A hearing device, e.g. a hearing aid, is configured to be worn by a user at or in an ear or to be fully or partially implanted in the head at an ear of the user. The hearing device comprises a) an input unit for providing at least one electric input signal in a time frequency representation k, m, where k and m are frequency and time indices, respectively, and k represents a frequency channel, the at least one electric input signal being representative of sound and comprising target signal components and noise components; and b) a signal processor comprising b1) a target signal estimator for providing an estimate of the target signal; b2) a noise estimator for providing an estimate of the noise; b3) a gain estimator for providing respective gain values in said time frequency representation in dependence of said target signal estimate and said noise estimate, wherein said gain estimator comprises a neural network, wherein the weights of the neural network have been trained with a plurality of training signals, and wherein the outputs of the neural network comprise real or complex valued gains, or separate real valued gains and real valued phases. The invention may e.g. be used in audio devices, such as hearing aids, headsets, mobile telephones, etc., operating in noisy acoustic environments.