Nonlinear Gain Smoothing for Noise Suppressor Musical Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional noise suppressors in communication systems often introduce 'musical' artifacts due to inappropriate gain application in certain frequency bands, leading to inadequate noise suppression or amplification, especially in noisy environments.
Innovation Solution
An enhanced noise suppressor that employs a frequency-domain conversion using FFT, DFT, or DCT, followed by adaptive gain curve calculation and nonlinear post-filtering, which adjusts gains based on signal and noise power estimates, and applies a smoothed gain curve to reduce noise fluctuations across frequency bands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional noise suppressors apply band-specific gain to suppress noise in frequency bands, then noise suppression performance is improved, but musical artifacts are introduced due to inappropriate gain application
Solution Approach 1:
The patent implements dynamic gain adjustment by continuously adapting the gain curve based on real-time noise power estimates and signal conditions. The gain for each frequency band is dynamically modified frame-by-frame rather than using fixed gains, allowing the system to respond to changing acoustic environments and avoid static gain-induced artifacts.
Solution Approach 2:
The patent changes the gain parameter adaptively based on noise power estimates and signal-to-noise ratio conditions. By modifying the gain curve parameters dynamically according to current acoustic conditions, the system optimizes noise suppression while minimizing artifacts. The gain application is adjusted based on multiple parameters including noise power, signal power, and estimated speech presence.
2Reliability
If noise suppressors suppress noise in certain frequency bands, then noise reduction is achieved, but intelligibility of voice signals may be impaired due to excessive suppression
Solution Approach 1:
The patent applies different gain values to different frequency bands based on local noise characteristics and speech content. Each frequency band is processed independently with its own gain adjustment, allowing aggressive suppression in noise-dominated bands while preserving gains in speech-rich bands. This localized approach maintains voice intelligibility while achieving effective noise reduction.
Solution Approach 2:
The system uses feedback from noise power estimates, signal power estimates, and speech detection results to continuously adjust the gain curve. The gain application in each frequency band is determined by feedback from the current acoustic conditions, ensuring that suppression levels are optimized to preserve speech intelligibility while reducing noise.
3Reliability
If noise suppressors apply high gain to suppress noise, then noise suppression is improved, but noise fluctuations are amplified creating artifacts
Solution Approach 1:
The patent implements dynamic gain curve adaptation that responds to changing noise conditions while maintaining stability. The gain curve is continuously updated based on noise power estimates and signal conditions, allowing the system to adapt to varying noise levels without introducing instability or artifacts from abrupt gain changes.
Data Source
AI summary
A noise suppressor has a band extractor to separate signal by frequency band; and per-band units for each of band including noise estimator and SNR computation units. The per-band unit has a histogrammer to give histograms of current and past SNRs, and a gain-curve updater computes gain curves from the histogram. Gain curves are used to determine raw gains from current SNRs, raw gain is filtered and controls a variable gain unit to provide band-specific gain-adjusted, signals that are recombined into a noise-reduced frequency-domain output. Raw gain filtering may include finite-impulse-response filtering and weighted averaging of intermediate gains of a current and adjacent-band per-band unit. The method includes separating an input into frequency bands, estimating in-band noise, and deriving a band SNR. Then, histogramming the SNR and updating a gain curve from the histogram, and finding a raw gain using the gain curve and current SNR.