Hearing Aid Neural Encoding for Low-Latency Frequency Resolution

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing hearing aids suffer from high latency in signal processing, which affects the effectiveness of hearing compensation and sound perception.

Innovation Solution

A hearing aid design utilizing a neural network-based encoder and decoder system that converts audio signals between domains, optimizing parameters to minimize latency and enhance hearing compensation, allowing for high-frequency resolution and low-latency sound processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional signal processing methods are used in hearing aids, then frequency resolution can be improved, but processing latency increases

Engineering Contradiction:
Improvefrequency resolutionVSAvoidprocessing latency
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The audio signal is divided into overlapping frames of N1 samples, with each frame processed independently through the encoder-neural network-decoder pipeline. This segmentation allows parallel processing of multiple frames, reducing overall latency while maintaining frequency resolution through the transform domain analysis of each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The encoder transforms N1 time-domain samples into N2 frequency-domain coefficients (where N2 > N1), transitioning from time domain to frequency domain representation. This dimensional transformation enables high frequency resolution without requiring long processing windows, as the frequency information is extracted directly through the encoder transform rather than through long FFT windows that would increase latency.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Measurement precision

If the number of frequency bins is increased for better frequency resolution, then processing complexity and latency increase

Engineering Contradiction:
Improvefrequency resolutionVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system dynamically adjusts the number of frequency bins N2 based on the specific hearing loss profile of the user. The encoder and neural network are trained together with a custom cost function that optimizes performance for the individual's auditory characteristics, allowing the frequency resolution to be adapted to actual needs rather than using a fixed high number of bins for all users.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter N2 (number of frequency bins) from a fixed large value to an optimized value determined by training. The encoder transform and neural network parameters are jointly optimized to achieve the required frequency resolution with minimal N2, reducing processing complexity while maintaining the necessary frequency discrimination capability.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If conventional encoder-decoder transforms are used, then frequency analysis is improved, but time domain reconstruction accuracy deteriorates

Engineering Contradiction:
Improvefrequency analysisVSAvoidtime domain reconstruction accuracy
Core Design Contradiction:
Measurement precisionVSManufacturing precision

Solution Approach 1:

The system uses a trained neural network that receives the encoded frequency-domain coefficients and produces enhanced frequency coefficients. The decoder then transforms these back to the time domain. The neural network is trained with a cost function that directly optimizes time-domain reconstruction accuracy, using feedback from the decoded output to adjust the frequency domain processing parameters and ensure accurate time-domain reconstruction.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent replaces the conventional fixed mathematical transform (such as standard FFT-based encoder-decoder pairs) with a learned transform implemented through trained neural networks. The encoder and decoder transforms are not fixed mathematical operations but are adapted through training data to optimize both frequency analysis and time-domain reconstruction, substituting rigid mechanical transform relationships with flexible learned mappings.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4099724B1A low latency hearing aid
Publication Date: 2026.01.21 OTICON
  • EP4099724B1 patent drawingFigure 1
  • EP4099724B1 patent drawingFigure 2
  • EP4099724B1 patent drawingFigure 3A

AI summary

A hearing aid comprises at least one input unit for providing at least one stream of samples of an electric input signal in a first domain; at least one encoder configured to convert said at least one stream of samples of the electric input signal in the first domain to at least one stream of samples of the electric input signal in a second domain; a processing unit configured to process said at least one electric input signal in the second domain, to provide a compensation for the user's hearing impairment, and to provide a processed signal as a stream of samples in the second domain; a decoder configured to convert said stream of samples of the processed signal in the second domain to a stream of samples of the processed signal in the first domain. The at least one encoder is configured to convert a first number of samples from said at least one stream of samples of the electric input signal in the first domain to a second number of samples in said at least one stream of samples of the electric input signal in the second domain. The decoder is configured to convert said second number of samples from said stream of samples of the processed signal in the second domain to said first number of samples in said stream of samples of the electric input signal in the first domain. The second number of samples is larger than the first number of samples. The at least one encoder is trained, and at least a part of said processing unit providing said compensation for the user's hearing impairment is implemented as a trained neural network. A method of operating a hearing aid is further disclosed. Thereby an improved hearing aid may be provided.