Source Separation for Hearing Aid Audio Dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Hearing assistance devices face challenges in processing audio signals with multiple sources of different levels, leading to improper gain application and resulting in artificial envelope fluctuations, distorted sound quality, and impaired sound localization, especially in complex audio mixes like music and movie soundtracks.
Innovation Solution
The implementation of source separation and remixing techniques in an external streaming audio device, which isolates individual audio sources, applies optimal gain processing, and mixes them back into the original signal to enhance sound quality while minimizing artifacts, using methods like non-negative matrix factorization for source separation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If dynamic range compression is applied to mixed audio signals in hearing assistance devices, then the loudness normalization function is achieved, but artificial envelope fluctuations and cross-modulation artifacts occur when multiple simultaneous sounds are present
Solution Approach 1:
The patent segments the audio signal processing into multiple frequency channels, with each channel independently applying dynamic range compression. This allows different gain values to be applied to different frequency regions simultaneously, preventing cross-modulation artifacts while maintaining loudness normalization for each frequency band.
Solution Approach 2:
The patent implements time-varying gain control with separate attack and release time constants for each frequency channel. The compression parameters dynamically adapt to the instantaneous energy levels in each channel, allowing the system to respond differently to transient versus sustained sounds and reducing artificial envelope fluctuations.
2Device complexity
If a single gain value is applied to mixed sounds in the same frequency region, then the compression algorithm is simplified, but individual sound sources receive improper gain and distorted loudness perception
Solution Approach 1:
The patent transforms the single-dimension gain control problem into a multi-dimensional solution by introducing frequency channel separation. Each frequency channel operates as an independent dimension with its own gain control, allowing precise gain application to different sound sources based on their spectral characteristics while maintaining manageable algorithmic complexity.
3Object-generated harmful factors
If multiband compression is used to reduce cross-modulation artifacts, then frequency-specific gain control is improved, but the device complexity and processing requirements increase
Solution Approach 1:
The patent implements a moderate number of frequency channels (e.g., 4-8 bands) rather than attempting to separate every frequency component. This partial action approach reduces cross-modulation artifacts sufficiently while keeping the processing complexity and device requirements within practical limits for hearing assistance devices.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This application relates to a systems and methods for enhanced dynamics processing of streaming audio by source separation and remixing for hearing assistance devices, according to one example. In one embodiment, an external streaming audio device processes sources isolated from an audio signal using source separation, and mixes the resulting signals back into the unprocessed audio signal to enhance individual sources while minimizing audible artifacts. Variations of the present system use source separation in a side chain to guide processing of a composite audio signal.