Adaptive Level Estimation for Speech Intelligibility in Hearing Aids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio processing devices, such as hearing aids, face challenges in accurately estimating sound levels due to the influence of stationary narrowband noise, which can mask low-level signal content like speech, leading to decreased intelligibility and inappropriate amplification.
Innovation Solution
A hybrid level estimation approach is proposed, utilizing a high-resolution level estimator in many frequency bands for slowly varying signals and a low-resolution estimator in few frequency bands for fast varying signals, with a fading scheme to combine the estimates, ensuring proper sound perception and speech intelligibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single level estimator is used for all frequency bands, then device complexity is reduced, but measurement precision deteriorates due to inability to distinguish between stationary noise and dynamic speech signals
Solution Approach 1:
The patent divides the level estimation task into two separate estimators: a first level estimator for stationary noise assessment and a second level estimator for dynamic speech detection. This segmentation allows each estimator to be optimized for its specific function, improving overall measurement precision without requiring a single complex estimator to handle all scenarios
Solution Approach 2:
The patent implements dynamic switching between different estimation approaches based on signal characteristics. The system adapts its estimation strategy by selecting between slow adaptation (for stationary noise) and fast adaptation (for dynamic speech), allowing the device to respond appropriately to different acoustic conditions without increasing structural complexity
2Speed
If fast adaptation is used for level estimation, then responsiveness to speech changes is improved, but noise influence increases due to premature reaction to noise fluctuations
Solution Approach 1:
The patent applies different adaptation characteristics to different estimation tasks. The first level estimator uses slow adaptation with larger time constants to assess stationary noise levels, while the second level estimator uses fast adaptation with smaller time constants to detect dynamic speech. This local differentiation of adaptation quality allows fast response to speech without premature reaction to noise
Solution Approach 2:
The patent introduces an intermediary mechanism (the first level estimator) that provides a stable noise reference. This intermediary allows the second fast-adapting estimator to distinguish between genuine speech signals and noise fluctuations by comparing against the stable noise assessment, enabling fast adaptation without increased noise influence
3Stability of the object's composition
If slow adaptation is used for level estimation, then noise stability is improved, but speech intelligibility deteriorates due to delayed response to speech onsets
Solution Approach 1:
The patent segments the adaptation function into two parallel estimators with different time constants. The first estimator provides stable, slow adaptation for noise assessment, while the second estimator provides fast adaptation for speech detection. This segmentation allows the system to maintain noise stability without sacrificing speech detection accuracy
Solution Approach 2:
The patent merges the outputs of two estimators with complementary characteristics. By combining the stable noise assessment from the first estimator with the responsive speech detection from the second estimator, the system achieves both noise stability and speech detection accuracy simultaneously
4Measurement precision
If high-resolution level estimation in many frequency bands is implemented, then speech intelligibility is improved, but computational load increases
Solution Approach 1:
The patent segments the frequency spectrum into multiple bands and applies different estimation resolutions to different bands based on their characteristics. This allows high-resolution estimation where speech information is present while using lower resolution where only stationary noise exists, reducing overall computational energy while maintaining speech intelligibility
Data Source
AI summary
An adaptive level estimator for providing a level estimate of an electric input signal representing sound is provided. The adaptive level estimator comprises a first level estimator that provides a first level estimate of the electric input signal in a first number K1 of frequency bands; a second level estimator that provides attack/release time constants associated with a second level estimate of the electric input signal in a second number K2 of frequency bands, wherein K2 is smaller than K1; and a level control unit that provides a resulting level estimate based on said first level estimates and said attack/release time constants associated with said second level estimates. The level estimator may be used in devices or applications that benefit from a dynamic adaptation of an input signal level to a listener's dynamic range of sound level perception, or to any other specific dynamic range deviating from that of the environment sound.