Neural Audio Separation in Hearing Devices for Complex Sound Scenes

Solution Overview

Problem

Existing signal processing devices struggle to efficiently and flexibly process audio signals, particularly in complex environments with multiple sound sources, leading to information loss and high computational complexity.

Innovation Solution

A signal processing device utilizing a neural network that can dynamically adjust the number of audio signals separated, allowing for adaptive customization based on the environment, with sequential neural networks for conditioning and separation, enabling efficient and precise audio signal processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If the number of separated audio signals is increased to capture all sound sources in complex environments, then information completeness is improved, but computational complexity increases

Engineering Contradiction:

Improveinformation completenessVSAvoidcomputational complexity

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

The neural network dynamically adjusts the number of separated audio signals based on environmental complexity. In simple environments, fewer signals are separated to reduce computational load, while in complex environments with multiple sound sources, the network increases the number of separated signals to capture all relevant information, thus adaptively resolving the contradiction between information completeness and computational complexity

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of the number of separated audio signals according to environmental conditions. By monitoring acoustic scene complexity and adjusting the separation count accordingly, the system maintains information completeness when needed while reducing computational complexity in simpler scenarios

Inventive Principle:

Principle #35Parameter changes

2Loss of information

If the number of separated audio signals is increased to capture all sound sources, then information completeness is improved, but power consumption increases

Engineering Contradiction:

Improveinformation completenessVSAvoidpower consumption

Core Design Contradiction:

Loss of informationVSUse of energy by moving object

Solution Approach 1:

The neural network dynamically adjusts the number of separated audio signals based on environmental complexity. In simple environments, fewer signals are separated to reduce computational load and power consumption, while in complex environments with multiple sound sources, the network increases the number of separated signals to capture all relevant information, thus adaptively resolving the contradiction between information completeness and power consumption

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of the number of separated audio signals according to environmental conditions. By monitoring acoustic scene complexity and adjusting the separation count accordingly, the system maintains information completeness when needed while reducing power consumption in simpler scenarios

Inventive Principle:

Principle #35Parameter changes

3Device complexity

If the number of separated audio signals is reduced in simple situations, then computational complexity is reduced, but adaptability decreases

Engineering Contradiction:

Improvecomputational complexityVSAvoidadaptability

Core Design Contradiction:

Device complexityVSAdaptability or versatility

Solution Approach 1:

The neural network dynamically adjusts the number of separated audio signals based on environmental complexity. In simple environments, fewer signals are separated to reduce computational load, while in complex environments with multiple sound sources, the network increases the number of separated signals to capture all relevant information, thus adaptively resolving the contradiction between information completeness and computational complexity

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes the parameter of the number of separated audio signals according to environmental conditions. By monitoring acoustic scene complexity and adjusting the separation count accordingly, the system maintains information completeness when needed while reducing computational complexity in simpler scenarios

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20260073934A1Signal processing device, system and method for processing audio signals

Publication Date: 2026.03.12 SONOVA AG

AI summary

A hearing device for processing audio signals is described. The hearing device has an input interface for receiving an input signal, a neural network for separating one or more audio signals from the input signal, and an output interface for outputting an output signal, wherein the output signal comprises the one or more audio signals separated from the input signal. The hearing device is configured to determine a priority parameter for each of the one or more audio signals, modulate each audio signal in the one or more audio signals based on the re-spective priority parameter, combine the one or more audio signals to produce an output signal, and output the output signal via the output interface.