Wind Noise Suppression Using Neural Gain Mixing in Audio Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing audio processing solutions for wind noise reduction fail to effectively suppress wind noise when it is present in only one audio signal out of two, leading to residual noise and reintroduction of noticeable wind noise due to aggressive noise processing techniques.
Innovation Solution
A method involving a wind noise suppressor module with a high-pass filter and a neural network to predict gains for noise reduction, combined with a mixer to generate an output audio signal, while using a wind noise indicator to control the processing dynamically, ensuring minimal distortion and preserving desired audio content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If aggressive noise processing techniques are used to suppress wind noise, then wind noise reduction is improved, but audio distortion and reintroduction of wind noise increase
Solution Approach 1:
The audio signal processing is divided into separate channels: a first channel processed through aggressive neural network-based noise suppression, and a second channel processed through conservative wind noise suppressor. This segmentation allows each channel to handle different aspects of noise reduction independently, preventing the reintroduction of wind noise while maintaining audio quality.
Solution Approach 2:
Different processing qualities are applied to different channels based on their specific characteristics. The first channel uses aggressive processing when needed, while the second channel uses conservative processing to preserve audio fidelity. This local differentiation of processing quality resolves the contradiction between effective noise suppression and audio distortion.
2Device complexity
If wind noise suppressor processes both audio signals equally, then processing simplicity is maintained, but residual noise remains when wind noise is present in only one signal
Solution Approach 1:
The system applies different processing strategies to different channels based on wind noise detection. When wind noise is detected in only one channel, that channel receives targeted processing while the other channel maintains its original conservative processing. This local adaptation eliminates residual noise without significantly increasing overall system complexity.
Solution Approach 2:
The processing approach dynamically adapts based on real-time wind noise detection results. The system transitions between different processing modes (aggressive vs. conservative) depending on the detected noise conditions, allowing optimal noise suppression while maintaining simplicity under normal conditions.
3Reliability
If neural network processing is applied to both audio signals, then noise reduction consistency is improved, but spatial audio object positioning accuracy decreases
Solution Approach 1:
The system applies neural network-based aggressive processing selectively to the first audio signal channel, while applying conservative processing to the second channel. This localized application maintains noise reduction consistency where needed while preserving spatial audio object positioning accuracy in the other channel.
Solution Approach 2:
By segmenting the processing into separate channels with different processing characteristics, the system maintains consistency in noise reduction for the processed channel while preserving spatial accuracy for the unprocessed channel, thus resolving the contradiction between consistency and precision.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The method effectively suppresses wind noise without introducing unwanted distortions, maintaining the desired audio content and ambiance, and adaptively responding to varying wind noise conditions.
Implementation Method 1
the wind noise suppressor module comprising a high-pass filter
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
The present disclosure relates to a method and system (1) for suppressing wind noise. The method comprises obtaining an input audio signal (100, 100') comprising a plurality of consecutive audio signal segments (101, 102, 103, 101', 102', 103') and suppressing wind noise in the input audio signal with a wind noise suppressor module (20) to generate a wind noise reduced audio signal. The method further comprises sing a neural network (10) trained to predict a set of gains for reducing noise in the input audio signal (100, 100') given samples of the input audio signal (100, 100'), wherein a noise reduced audio signal is formed by applying said set of gains to the input audio signal (100, 100') and mixing the wind noise reduced audio signal and the noise reduced audio signal with a mixer (30) to obtain an output audio signal with suppressed wind noise.