Hearing Audio Processing with Scene-Specific Noise Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing audio data processing methods struggle to effectively enhance speech intelligibility in noisy environments, despite advancements in voice detection and acoustic scene classification, necessitating improved methods and devices for clearer speech presentation.

Innovation Solution

A computer-implemented method involving a speech removal module, acoustic scene classification, and specialized noise reduction modules, utilizing pre-trained neural networks tailored to specific acoustic scenes, combined with Time-Frequency masking to efficiently separate speech components and reduce background noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If voice detection is used to amplify speech frequency components, then speech intelligibility is improved, but the system cannot adapt to different acoustic environments effectively

Engineering Contradiction:
Improvespeech intelligibilityVSAvoidacoustic environment adaptation
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system dynamically adapts to different acoustic environments by using acoustic scene classification to identify the current environment type (e.g., train, café, street) and then selecting or adjusting noise reduction parameters accordingly. This allows the speech enhancement system to maintain high intelligibility across varying acoustic conditions rather than using fixed amplification parameters.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes processing parameters based on the identified acoustic scene. Different acoustic environments have different noise characteristics, and the system adjusts noise reduction strength, frequency weighting, and other parameters according to the scene classification result, thereby optimizing speech intelligibility for each specific environment.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If acoustic scene classification is used to identify environment type, then adaptability to different environments is improved, but computational complexity increases

Engineering Contradiction:
Improveacoustic environment classificationVSAvoidcomputational complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The acoustic scene classification problem is segmented into distinct environment categories (train, café, street, etc.), each with characteristic acoustic profiles. This segmentation allows the system to use simpler classification rules or pre-trained models for each category rather than attempting to classify all possible acoustic variations continuously, reducing overall computational complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system performs preliminary acoustic scene classification to identify the general environment type before applying detailed noise reduction processing. This preliminary classification allows subsequent processing to be optimized based on the known scene type, reducing the computational burden of more complex real-time analysis.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If speech components are removed to determine acoustic scene, then scene classification accuracy is improved, but speech content is lost

Engineering Contradiction:
Improveacoustic scene classification accuracyVSAvoidspeech content
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The audio signal is segmented into speech components and background sound components. By separating these components, the system can analyze the background sound for acoustic scene classification without the speech content interfering with the classification accuracy, while simultaneously preserving the speech content for later enhancement and output.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The background sound serves as an intermediary for acoustic scene classification. Instead of classifying the mixed audio signal directly, the system extracts and uses only the background sound portion as the intermediary for scene identification, thereby improving classification accuracy without permanently losing speech content.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP4300491B1A method for transforming audio input data into audio output data and a hearing device thereof
Publication Date: 2025.11.12 GN HEARING AS
  • EP4300491B1 patent drawingFigure 1
  • EP4300491B1 patent drawingFigure 2

AI summary

A computer-implemented method (200) for transforming audio input data (110) into audio output data (130) is provided. The method (200) comprises receiving (202) audio input data (110), providing (204) background sound data (122) by separating speech components from the audio input data (110) by using a speech removal module (120), determining (206) acoustic scene data (126), linked to an acoustic scene (AS) matching the background sound data (122), by using an acoustic scene classifier (ASC) module (124), selecting (208) a specialized noise reduction (S-NR) module (128A-C) based on the acoustic scene data (126), and processing (210) the audio input data (110) by using the specialized noise reduction (S-NR) module (128A-C) such that the audio output data (130) is generated.