Sound Capture Control for Wind Noise Suppression Without Tone Loss
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional wind noise suppression techniques, such as high-pass filtering and spectral subtraction, often result in tone alteration and loss of target sound components, especially when wind noise levels are high, and existing methods struggle to accurately restore sound quality without distorting the target sound.
Innovation Solution
A sound capture apparatus that includes an estimation unit for noise signal estimation, a detection unit for noiseless state detection, and a learning unit for analyzing and modeling target sound characteristics, allowing for precise restoration of target sounds while suppressing noise without altering tone or losing sound components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If high-pass filter is used to suppress wind noise, then wind noise suppression is improved, but target sound quality deteriorates due to tone alteration
Solution Approach 1:
The patent segments the audio signal into multiple frequency bands (low-band, middle-band, high-band) and processes each band separately. The low-band signal undergoes wind noise suppression, while the middle and high-band signals are used to generate restoration signals that compensate for the suppressed low-frequency components, thereby maintaining target sound quality while removing wind noise.
Solution Approach 2:
The patent dynamically adjusts the suppression amount of the high-pass filter based on the estimated wind noise level. When wind noise is detected, the suppression amount is increased; when wind noise is absent, the suppression amount is reduced or eliminated. This adaptive parameter adjustment prevents unnecessary tone alteration while maintaining effective wind noise suppression.
2Object-affected harmful factors
If spectral subtraction is used to suppress wind noise, then wind noise suppression is improved, but target sound components are lost
Solution Approach 1:
The patent uses the middle-band and high-band signals (which contain target sound information) as feedback to generate restoration signals that are added back to the low-band signal. This feedback mechanism ensures that target sound components are preserved and restored after wind noise suppression, preventing the loss of important audio information.
Solution Approach 2:
The patent performs preliminary analysis of the captured audio signal to estimate wind noise characteristics and identify noiseless intervals before applying suppression. By learning the target sound characteristics during noiseless intervals in advance, the system can accurately restore target sound components after suppression without losing them.
3Object-affected harmful factors
If wind noise suppression processing is applied, then wind noise is reduced, but target sound component restoration is inaccurate
Solution Approach 1:
The patent performs preliminary learning of target sound characteristics during noiseless intervals before wind noise occurs. The learning unit analyzes the captured audio signal during these intervals to build a model of the target sound, which is then used to accurately restore target sound components after wind noise suppression, significantly improving restoration accuracy.
Solution Approach 2:
The patent applies different processing strategies to different frequency bands and different time intervals. The low-band signal receives wind noise suppression, while the middle and high-band signals are used for restoration. During noiseless intervals, the system learns target sound characteristics; during wind noise intervals, it applies suppression and restoration. This localized quality approach maximizes restoration accuracy.
Data Source
AI summary
A noise signal is estimated based on a captured audio signal captured from a sound capture unit. It is determined whether the estimated noise signal thus estimated is in a noiseless state. If it is determined that the estimated noise signal is in the noiseless state, the captured audio signal is analyzed as a target sound signal, and a characteristic obtained by the analysis is learned and modeled, thereby generating a target sound model.