Binaural Hearing Device Common-Gain Filter for Noise Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional binaural hearing devices face challenges in achieving significant noise reduction while preserving speech quality and spatial cues, often resulting in speech distortion and processing artifacts like 'musical noise', which requires heuristic adjustments and critical estimation of local time-varying a-priori SNR.
Innovation Solution
A hearing device and method that determine a common-gain filter using statistical information of pre-evaluated SNR probability distributions, replacing a-priori SNR estimation with Bayesian SNR marginalization, and applying binaural Bayesian filters to both ear signals for improved noise reduction and speech quality, avoiding heuristic adjustments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional noise reduction methods (spectral subtraction, Wiener filter, MMSE STSA estimator) are applied independently to each ear signal, then noise reduction performance is improved, but spatial cues are destroyed
Solution Approach 1:
The patent merges the processing of both ear signals by applying a common gain filter derived from combined statistical information of both channels. Instead of independent processing, the system combines the noise estimates and SNR information from both ears to determine a single common gain filter that is applied to both ear signals, thereby preserving spatial cues while achieving noise reduction.
2Loss of information
If a common-gain filter is applied to both ear signals to preserve spatial cues, then spatial cue preservation is improved, but noise reduction performance deteriorates
Solution Approach 1:
The patent changes the parameters used in common-gain filtering by incorporating statistical information about SNR probability distributions. Instead of using simple instantaneous SNR estimates, the system uses pre-evaluated statistical information and current SNR estimates to determine optimized common gain filters that achieve both spatial cue preservation and improved noise reduction performance.
3Object-affected harmful factors
If local time-varying a-priori SNR estimation is performed to enable common-gain filtering, then noise reduction is improved, but computational complexity and need for heuristic adjustments increase
Solution Approach 1:
The patent performs preliminary action by pre-evaluating and storing statistical information about SNR probability distributions before actual speech enhancement is needed. This pre-computed statistical information is then used during runtime to quickly determine common gain filters without requiring complex real-time SNR estimation algorithms or heuristic adjustments, thereby reducing computational complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A hearing device (200) being configured to: receive a first input signal (201) and a second input signal (202) (401); determine a first ear signal (203) based on the first input signal (201) and determine a second ear signal (204) based on the second input signal (202) (402); obtain statistical information (205) indicative of a pre-evaluated SNR probability distribution that depends on a global SNR (403); estimate a current SNR (206) based on the first input signal (201) and the second input signal (202) (404); determine one or more common-gain filters (207), wherein the one or more common-gain filters (207) are determined based on the statistical information (205) and the estimated current SNR (206) (405); and apply the one or more common-gain filters (207) to both the first ear signal (203) and the second ear signal (204) (406).