Wind Noise Detection Using Multi-Microphone Cross-Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wind noise significantly degrades audio quality in mobile devices, particularly in outdoor settings, due to its broadband and non-stationary nature, making it challenging to detect and remove effectively.
Innovation Solution
A method and system using multiple microphones to detect wind noise by filtering audio signals to reduce higher and lower frequencies, computing cross-correlation at multiple delays, and comparing the maximum cross-correlation with a threshold to identify wind noise presence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If wind noise detection is implemented using traditional single-microphone methods, then the detection capability is limited, but the device complexity remains low
Solution Approach 1:
The patent divides the audio signal processing into multiple frequency bands (low-frequency, mid-frequency, high-frequency components) and processes each band separately through different filtering paths. This segmentation allows the system to detect wind noise characteristics in specific frequency ranges while maintaining manageable complexity in each processing branch.
Solution Approach 2:
The patent transitions from single-microphone temporal analysis to multi-microphone spatial-temporal analysis by computing cross-correlation between signals from multiple microphones at multiple delay values. This adds a spatial dimension (inter-microphone time differences) to the detection process, significantly improving wind noise identification accuracy.
2Measurement precision
If multiple audio sources are used for wind noise detection, then the detection accuracy improves, but the processing complexity increases
Solution Approach 1:
The patent applies different processing characteristics to different frequency components: low-pass filtering for low-frequency wind noise components, band-pass filtering for mid-frequency components, and high-pass filtering for high-frequency components. Each frequency band is processed with appropriate characteristics to optimize wind noise detection while minimizing unnecessary processing complexity.
Solution Approach 2:
The patent performs preliminary filtering of audio signals into different frequency bands before cross-correlation computation. By pre-processing the signals to isolate wind noise-dominated frequency ranges, the subsequent cross-correlation analysis becomes more efficient and focused, reducing the overall processing complexity.
3Reliability
If cross-correlation is computed at multiple delays, then the wind noise detection reliability improves, but the computational load increases
Solution Approach 1:
The patent computes cross-correlation at a limited set of discrete delay values rather than continuously across all possible delays. By selecting only the most relevant delay values (corresponding to expected wind noise propagation time differences between microphones), the system achieves sufficient detection reliability with reduced computational effort.
Solution Approach 2:
The patent changes the processing parameters by applying different filter cutoff frequencies to separate frequency bands before cross-correlation. This parameter transformation allows the system to focus computational resources on frequency ranges where wind noise is most prominent, improving reliability while managing power consumption.
Data Source
AI summary
A method for detecting wind noise is described. At least two audio signals are received. The at least two audio signals are filtered to reduce higher frequencies and to reduce lower frequencies to provide at least two filtered audio signals. The cross correlation of the at least two filtered audio signals is computed for multiple delays. A maximum cross correlation is determined from the cross correlations computed for the multiple delays. Wind noise is detected by comparing the maximum cross correlation with a threshold.