Ear-worn Device Noise Reduction via Adaptive Mixing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing ear-worn devices, such as hearing aids, face challenges in effectively reducing noise from interfering speakers and achieving directional sound enhancement due to limitations in beamforming and noise reduction techniques, especially in real-world environments.

Innovation Solution

The implementation of a combination of classical beamforming and noise reduction techniques, utilizing spatial information to preferentially attenuate interfering speakers and adaptively control noise mixing after speech and noise separation using neural networks.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Object-affected harmful factors

If conventional beamforming is used to attenuate interfering speakers, then some noise reduction is achieved, but the directionality pattern warps due to interference from the wearer's ear, head, and torso, limiting noise reduction to only a few dB

Engineering Contradiction:

Improvenoise from interfering speakersVSAvoiddirectionality pattern stability

Core Design Contradiction:

Object-affected harmful factorsVSReliability

Solution Approach 1:

The patent segments the noise reduction process into multiple independent components: neural network-based speech-noise separation, spatial information extraction, and adaptive mixing control. This segmentation allows each component to be optimized independently, with the neural network handling spectral separation and the spatial processing handling directional information, thereby maintaining directionality stability while achieving effective noise reduction.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces spatial information as an intermediary between the raw audio signal and the final output. This intermediary carries directional characteristics that are extracted from the audio signal and used to guide the noise reduction process, ensuring that the directionality pattern remains stable and accurate even in the presence of body interference.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If beamforming steers nulls toward interfering speakers, then noise reduction is improved, but the steering is dominated by ambient noise in loud environments, reducing effectiveness

Engineering Contradiction:

Improveinterfering speaker attenuationVSAvoidnull steering accuracy

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

The patent performs preliminary speech-noise separation using a neural network before attempting spatial null steering. By pre-separating the speech and noise components in the spectral domain, the system obtains cleaner spatial information that is not dominated by ambient noise, enabling more accurate null steering toward interfering speakers.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements adaptive mixing control that uses feedback from the separated speech and noise components to dynamically adjust the mixing ratio. This feedback mechanism ensures that the noise reduction process adapts to changing environmental conditions and maintains optimal performance even when ambient noise levels vary.

Inventive Principle:

Principle #23Feedback

3Object-affected harmful factors

If neural networks separate speech from noise, then noise reduction is enhanced, but adaptive control of noise mixing is needed to maintain natural sound quality

Engineering Contradiction:

Improvenoise component removalVSAvoidnoise mixing control

Core Design Contradiction:

Object-affected harmful factorsVSEase of operation

Solution Approach 1:

The patent implements adaptive mixing control that automatically adjusts the noise mixing ratio based on spatial information and environmental conditions. The system self-regulates the amount of noise to be mixed back with the speech signal, eliminating the need for manual intervention while maintaining natural sound quality and optimal noise reduction performance.

Inventive Principle:

Principle #25Self-service

4Measurement precision

If spatial information-based noise reduction is applied, then directionality and SNR gain are improved, but device complexity increases due to multiple processing stages

Engineering Contradiction:

Improvesignal-to-noise ratio gainVSAvoidnoise reduction system structure

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent merges the neural network-based speech separation with spatial information processing in a unified framework. By combining these functions into an integrated system with shared computational resources and coordinated processing stages, the patent achieves high SNR gain and improved directionality while managing device complexity through efficient resource utilization and streamlined architecture.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS20250124946A1Ear-worn device providing enhanced noise reduction and directionality

Publication Date: 2025.04.17 FORTELL RESEARCH INC

AI summary

In some embodiments, an ear-worn device may include adaptive mixing control circuitry configured to control mixing circuitry to mix a speech component of an input audio signal with a noise component of the input audio signal based on a deviation of a short-term level of the noise component of the input audio signal from a long-term level of the noise component of the input audio signal. In some embodiments, an ear-worn device may include beamforming circuitry configured to generate a front audio signal and a back audio signal, and direction-of-arrival circuitry configured to bias an output audio signal based on a difference between the front audio signal and the back audio signal.