Dynamic Quantile Tracking for Noise Suppression in Hearing Aids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise suppression methods in hearing aids and speech communication devices face challenges with high computational complexity, large memory requirements, and the need for voice activity detection, which are not suitable for real-time processing and effective in handling non-stationary non-white noises.
Innovation Solution
A method using spectral subtraction with dynamic quantile tracking for noise spectrum estimation, which approximates quantile values without storing past spectral samples, allowing for frequency-bin dependent quantile estimation and reducing processing overheads, integrated with FFT-based analysis-synthesis for real-time noise suppression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If quantile-based noise estimation is used for spectral subtraction, then noise suppression effectiveness is improved, but computational complexity and memory requirements increase
Solution Approach 1:
The patent changes the parameter approach by using dynamic quantile tracking that adapts quantile values based on current signal conditions rather than using fixed quantile values. This allows the system to maintain high noise suppression effectiveness while reducing computational complexity through parameter adaptation rather than exhaustive calculation
Solution Approach 2:
The patent implements dynamic quantile tracking where quantile values are updated continuously based on current spectral conditions. This dynamic approach replaces static quantile-based estimation, allowing the system to track non-stationary noise effectively while maintaining lower computational requirements through incremental updates rather than full re-calculation
2Measurement precision
If spectral samples are stored for quantile estimation, then noise estimation accuracy is improved, but memory requirements increase
Solution Approach 1:
The patent implements a self-service mechanism where the quantile tracker maintains a compact internal state that is automatically updated with each new spectral frame. This eliminates the need to store large numbers of historical spectral samples while maintaining accurate noise estimation through the self-updating quantile state
Solution Approach 2:
The patent segments the noise estimation process into incremental quantile updates rather than requiring storage of complete spectral histories. By processing information in small, manageable quantile state updates, the system achieves accurate noise estimation with minimal memory requirements
3Reliability
If voice activity detection is used for noise estimation, then speech quality is improved, but processing delay and complexity increase
Solution Approach 1:
The patent extracts the voice activity detection step from the noise estimation process. By using quantile-based tracking that operates independently of VAD decisions, the system eliminates the processing delay and complexity associated with VAD while maintaining speech quality through continuous noise spectrum tracking
Solution Approach 2:
The patent performs preliminary noise spectrum tracking continuously without waiting for VAD to identify speech segments. The quantile-based tracker maintains ready-to-use noise estimates in advance, eliminating the delay that would otherwise occur when noise estimation must wait for VAD confirmation of non-speech segments
Data Source
AI summary
A method for speech enhancement in speech communication devices and more specifically in hearing aids for suppressing stationary and non-stationary background noise in the input speech signal signals is disclosed. The method uses spectral subtraction wherein the noise spectrum is updated using quantile-based estimation without voice activity detection and the quantile values are approximated by dynamic quantile tracking without involving large storage and sorting of past spectral samples. The technique permits use of a different quantile at each frequency bin for noise estimation without introducing processing overheads. The preferred embodiment uses analysis-modification-synthesis based on Fast Fourier transform (FFT) and it can be integrated with other FFT-based signal processing techniques used in the hearing aids and speech communication devices. A noise suppression system based on this method and using hardware with an audio codec and a digital signal processor chip with on-chip FFT hardware is also disclosed.