Hearing Aid Signal Processing Using SII Gradient Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hearing aids face challenges in optimizing speech intelligibility in real-time using limited processing resources, as they rely on predefined gain adjustments that do not account for inter-band interactions and require numerical optimization routines, which are complex to derive and implement effectively.
Innovation Solution
A method for processing signals in hearing aids that uses a speech and noise estimator to calculate optimized gain values for each frequency band, incorporating a listening comfort model and automatic gain control to maximize the Speech Intelligibility Index (SII) through iterative optimization, employing a gradient ascent method to adjust gain values in real-time, while considering both speech and noise levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If numerical optimization routines are used to maximize speech intelligibility, then speech intelligibility is improved, but device complexity increases
Solution Approach 1:
The patent changes the parameters being optimized from raw gain values to SII gradients. By computing the gradient of the Speech Intelligibility Index with respect to gain settings and optimizing these gradient values instead, the system achieves more efficient convergence to optimal speech intelligibility while reducing computational complexity. This parameter transformation allows the use of simpler gradient-based optimization methods rather than complex numerical routines.
2Measurement precision
If real-time optimization is implemented, then speech intelligibility is improved, but processing resources are exceeded
Solution Approach 1:
The patent extracts and pre-computes the SII gradient values, separating the computationally intensive gradient calculation from the real-time optimization loop. By extracting the gradient information and using it to guide simpler iterative optimization steps, the system reduces the processing burden during real-time operation while maintaining speech intelligibility improvement.
Solution Approach 2:
The system performs preliminary computation of SII gradients before the main optimization process. By calculating the gradient of speech intelligibility with respect to gain settings in advance and using these pre-computed gradients to guide the optimization, the patent reduces the computational resources required during real-time operation, enabling implementation on hearing aids with limited processing power.
3Measurement precision
If gain levels are increased to improve speech intelligibility, then speech understanding is improved, but listening comfort deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the SII gradient is continuously computed based on the current gain settings and speech intelligibility measurement. This gradient feedback guides the iterative optimization process to adjust gain levels in a controlled manner, increasing gain only when it improves speech intelligibility while avoiding excessive gain increases that would cause loudness discomfort. The feedback loop ensures balanced optimization of both intelligibility and comfort.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method of processing a signal in a hearing aid (50) comprising the steps of determining a gradient of a speech intelligibility measure using a closed form expression of the gradient and modifying the frequency dependent hearing aid gains in dependence on the calculated gradient. The invention also provides a hearing aid (50).