Neural Microphone Processing for Speech in Noisy Hearing Aids

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Hearing aids struggle to distinguish target sounds, such as speech, from environmental noise due to loss of frequency resolution in the cochlea, leading to difficulty in noisy environments.

Innovation Solution

A neural network-based hearing system processes microphone signals using nonlinear processing to enhance target sounds, improving signal-to-noise ratio (SNR) and speech intelligibility without substantial distortion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional binaural beamformers are used to process microphone signals, then the system structure is simple and computationally efficient, but the signal-to-noise ratio improvement is limited to theoretical maximums

Engineering Contradiction:

Improvesignal-to-noise ratio improvementVSAvoidsignal processing algorithm complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces traditional linear beamforming algorithms with a neural network-based nonlinear processing system. The neural network learns optimal nonlinear transformations from training data, enabling superior signal-to-noise ratio improvement beyond what linear methods can achieve, while the network is deployed as a fixed computational model that processes audio signals in real-time.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The invention changes the fundamental parameters of signal processing by transitioning from linear operations (traditional beamforming) to nonlinear operations (neural network transformations). The neural network adjusts its internal parameters during training to optimize performance, and these learned parameters are then used to process incoming audio signals, achieving enhanced SNR improvement.

Inventive Principle:

Principle #35Parameter changes

2Reliability

If nonlinear signal processing is applied to enhance target sounds, then the signal-to-noise ratio improves beyond theoretical limits, but audible distortion of the target signal may occur

Engineering Contradiction:

Improvesignal-to-noise ratio improvementVSAvoidsignal fidelity

Core Design Contradiction:

ReliabilityVSManufacturing precision

Solution Approach 1:

The neural network is trained using feedback from ground truth audio signals. During training, the network processes noisy input signals and compares its output to the clean target signals, adjusting its internal parameters to minimize distortion while maximizing SNR improvement. This feedback mechanism ensures the network learns transformations that enhance signals without introducing substantial audible distortion.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary training of the neural network using simulated or recorded audio data before deployment. During this preliminary training phase, the network learns the optimal nonlinear transformations to apply to real-world audio signals, pre-adjusting its parameters to achieve the desired balance between SNR improvement and signal fidelity in actual operation.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20260025623A1Hearing device with neural network-based microphone signal processing

Publication Date: 2026.01.22 STARKEY LABORATORIES INC

AI summary

A hearing system performs nonlinear processing of signals received from a plurality of microphones using a neural network to enhance a target signal in a noisy environment. In various embodiments, the neural network can be trained to improve a signal-to-noise ratio without causing substantial distortion of the target signal. An example of the target sound includes speech, and the neural network is used to improve speech intelligibility.