Two-Scale Microphone Array for Wind Noise Signal Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing acoustic detection systems face challenges in accurately estimating acoustic signal parameters like direction of arrival and waveform due to wind noise, especially in transient signal detection and on mobile platforms, where wind noise is highly non-stationary and correlated, leading to biased estimates and limited benefits from mechanical windscreens.
Innovation Solution
A two-scale array configuration with closely spaced microphones to increase wind noise correlation, using a mathematical model that separates wind noise and acoustic signals by fitting pressure pulse data to a parametric model including terms for wind noise and acoustic signals, allowing for improved signal-to-noise ratio without prior knowledge of wind noise correlation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If mechanical windscreens are used to reduce wind noise, then the overall measured level of pressure fluctuations due to wind noise is reduced, but the correlation among fluctuations due to wind noise biases the estimates of direction of arrival and waveform
Solution Approach 1:
The sensor array is divided into multiple subarrays with specific spatial configurations. Each subarray processes signals independently to estimate wind noise correlation, which is then used to correct the overall direction of arrival estimation, separating the harmful wind noise effect from the acoustic signal measurement
Solution Approach 2:
The system changes the spatial parameters of sensor arrangement by using non-uniform spacing and multiple subarrays with different configurations. This allows the system to operate in a parameter regime where wind noise correlation can be measured and compensated, improving measurement precision while maintaining wind noise reduction
2Power
If sensors are spaced far apart to avoid wind noise correlation, then spatial averaging through beamforming can enhance SNR, but biased estimates of signal parameters are produced when wind noise is correlated
Solution Approach 1:
The system performs preliminary estimation of wind noise correlation structure from the sensor data before conducting the main signal processing. This preliminary action allows the system to adjust subsequent beamforming operations to compensate for wind noise correlation, ensuring accurate signal parameter estimation while maintaining high SNR
Solution Approach 2:
The system uses feedback from the measured wind noise correlation to adjust the beamforming weights and signal processing parameters. This feedback mechanism ensures that the spatial averaging benefits are realized while correcting for the biases introduced by wind noise correlation
3Power
If time averaging is used to improve detection abilities for continuous wave signals, then SNR gains are achieved, but time averaging is generally ineffective for transient acoustic signal detection
Solution Approach 1:
The system dynamically adapts its processing strategy based on the temporal characteristics of the detected signals. For transient signals, it uses dynamic thresholding and short-time analysis rather than long-time averaging, maintaining effectiveness across different signal types while achieving SNR improvement where applicable
4Measurement precision
If modified beamforming techniques are used to reduce bias from correlated wind noise, then accurate signal parameter estimation is achieved, but it is relatively difficult to determine the correlation structure of wind noise before acquiring data
Solution Approach 1:
The system determines wind noise correlation structure directly from the acquired sensor data without requiring external measurements or complex pre-characterization equipment. The correlation structure is self-determined through statistical analysis of the recorded signals, simplifying the overall system while achieving accurate compensation
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A two-scale array for detecting wind noise signals and acoustic signals includes a plurality of subarrays each including a plurality of microphones. The subarrays are spaced apart from one another such that the subarrays are configured to detect acoustic signals, and the plurality of microphones in each subarray are located close enough to one another such that wind noise signals are substantially correlated between the microphones in each subarray.