Audio Gain Smoothing for Noise Reduction Without Musical Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise reduction methods for audio signals often produce artifacts such as speech distortion and musical noise due to estimation errors and non-stationarity of noise, leading to undesirable spectral peaks and gain fluctuations.
Innovation Solution
The method involves post-processing raw gains determined by input processing using delta gain smoothing and decision-directed gain smoothing techniques to generate post-processed gains, which are then applied to the audio signal to reduce noise and artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If input processing methods are used to determine gains for noise reduction, then noise can be reduced, but artifacts such as musical noise and speech distortion are produced due to estimation errors and gain fluctuations
Solution Approach 1:
The patent applies preliminary smoothing actions to the raw gains before they are applied to the signal. By pre-smoothing the gain values using temporal and spectral smoothing techniques, the method prevents the formation of spurious peaks and gain fluctuations that would otherwise cause musical noise and speech distortion artifacts in the output signal.
Solution Approach 2:
The patent introduces an intermediate processing stage between gain determination and signal enhancement. This intermediate stage includes smoothing operations that act as mediators to reduce the harmful effects of raw gain fluctuations. The smoothed gains serve as an intermediary representation that preserves noise reduction benefits while eliminating artifacts.
2Reliability
If aggressive noise reduction is applied to remove undesired signals, then noise levels decrease, but speech distortion increases due to over-suppression of desired signal components
Solution Approach 1:
The patent applies partial smoothing to the gains rather than complete smoothing. By using smoothing factors that are less than 1 and applying smoothing selectively in temporal and spectral domains, the method achieves partial suppression of noise while preserving enough of the original signal characteristics to avoid speech distortion. The smoothing is applied to the extent needed to reduce artifacts without over-suppressing desired components.
Data Source
AI summary
A method, an apparatus, and logic to post-process raw gains determined by input processing to generate post-processed gains, comprising using one or both of delta gain smoothing and decision-directed gain smoothing. The delta gain smoothing comprises applying a smoothing filter to the raw gain with a smoothing factor that depends on the gain delta: the absolute value of the difference between the raw gain for the current frame and the post-processed gain for a previous frame. The decision-directed gain smoothing comprises converting the raw gain to a signal-to-noise ratio, applying a smoothing filter with a smoothing factor to the signal-to-noise ratio to calculate a smoothed signal-to-noise ratio, and converting the smoothed signal-to-noise ratio to determine the second smoothed gain, with smoothing factor possibly dependent on the gain delta.