Multi-Channel Hearing Aid Source Localization via Cumulative Statistics

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing multi-channel hearing systems face challenges in accurately detecting and localizing multiple audio sources, particularly when sources are from the same direction, due to insufficient information in small time-frequency blocks.

Innovation Solution

A method that involves detecting and filtering signal components based on specific features before localization, using decision-directed filtering and cumulative statistics, to enhance the reliability of source localization by combining signal components and reducing noise, allowing for precise identification and tracking of sources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If small time-frequency blocks are used for source localization, then the processing complexity is reduced, but the measurement precision of source localization deteriorates

Engineering Contradiction:
Improveprocessing complexityVSAvoidsource localization precision
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent combines multiple small time-frequency blocks that contain signal components from the same audio source into a single cumulative spectrum. This merging process aggregates the information from multiple blocks, thereby improving the precision of source localization while maintaining the computational efficiency of processing smaller blocks individually before combination.

Inventive Principle:
Principle #5Merging (Combining)

2Loss of energy

If signal components from multiple sources are processed together, then the data transmission is reduced, but the reliability of source identification deteriorates when sources are from the same direction

Engineering Contradiction:
Improvedata transmission requirementsVSAvoidsource identification reliability
Core Design Contradiction:
Loss of energyVSReliability

Solution Approach 1:

The patent segments the audio signal into multiple time-frequency blocks and processes them individually through decision-directed filtering. Each block is analyzed separately to identify signal components, and only the essential localization information is transmitted. This segmentation allows reliable source identification even when sources are from the same direction, while minimizing data transmission requirements.

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If decision-directed filtering is applied to detect signal components, then the measurement precision of source localization is improved, but the device complexity increases

Engineering Contradiction:
Improvesource localization precisionVSAvoidsignal processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies decision-directed filtering as a preliminary action before source localization. Signal components are detected and filtered in advance based on their characteristics, which prepares the data for more accurate localization. This preliminary processing improves measurement precision while keeping the overall system complexity manageable by organizing the processing steps in a logical sequence.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2405673B1Method for localising an audio source and multi-channel audio system
Publication Date: 2018.08.08 SIVANTOS PTE LTD
  • EP2405673B1 patent drawingFigure 1
  • EP2405673B1 patent drawingFigure 2~3
  • EP2405673B1 patent drawingFigure 4

AI summary

The method involves detecting a signal of a pre-determined class, which is caused by an audio source, in a curve input signal (I) of a multi-channel, binaural hearing system. The audio source is localized (26) based on the detected curved signal (II), and the pre-determined class is provided as a language class. Signal portions are filtered from the input signal in dependent upon the detected signal. The localization is implemented using cumulative statistics (28) and a localization algorithm. Predetermined characteristics e.g. formant, of the input signal is determined, during detection. An independent claim is also included for a multi-channel hearing system that comprises a localization device.