Microphone Array Noise Suppression via Isotropy Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing microphone array noise suppression systems face challenges in accurately estimating noise field correlations, leading to incorrect beamformer output noise estimation, especially in dynamic environments, and require significant computational resources, making real-time processing difficult for large arrays.
Innovation Solution
A noise field isotropy model is used to estimate the correlation between sound field phases in a microphone array, employing a power spectral density (PSD) estimation method that calculates a single aggregated transfer function from pair-wise microphone PSD differences, reducing computational complexity and enabling real-time noise estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional noise suppression algorithms are used to accurately estimate noise field correlations, then noise suppression accuracy is improved, but computational complexity increases significantly
Solution Approach 1:
The patent changes the parameter representation by using power spectral density (PSD) differences instead of direct correlation estimation, and introduces a transfer function to model the relationship between microphone pairs. This parameter transformation reduces computational complexity while maintaining noise suppression accuracy through the equation H(i,j)(ω) = (Φxx,i(ω) - Φxx,j(ω)) / (|H_i(ω)|^2 + ε), where the transfer function captures the essential correlation information with fewer computations.
Solution Approach 2:
The patent extracts only the essential correlation information needed for noise suppression by using PSD differences between microphone pairs. Instead of computing full correlation matrices, it extracts the differential PSD components and processes them through a transfer function, significantly reducing the computational burden while retaining the critical noise field correlation characteristics.
2Measurement precision
If complex algorithms are used to isolate and reduce noise from speech, then noise suppression accuracy is improved, but processing speed decreases
Solution Approach 1:
The patent transforms the noise estimation problem into a parameter optimization task by using PSD differences and transfer functions. This allows for faster computation through closed-form solutions and iterative refinement, achieving both high accuracy and real-time processing speeds by avoiding computationally intensive operations while maintaining precise noise field correlation estimation.
Solution Approach 2:
The patent implements a dynamic approach where the transfer function is continuously updated based on current audio conditions. The system adapts to changing noise environments by recalculating PSD differences and transfer function parameters in real-time, enabling both accurate noise suppression and fast response to dynamic acoustic conditions.
3Measurement precision
If accurate noise field correlation estimation is performed in large microphone arrays, then noise suppression accuracy is improved, but computational resources required increase
Solution Approach 1:
The patent segments the large microphone array into multiple pairs of microphones, processing each pair independently through the transfer function. This segmentation divides the computational task into smaller, manageable units that can be processed in parallel, reducing the overall computational resources required while maintaining accurate noise field correlation estimation across the entire array.
Solution Approach 2:
The patent reduces computational resources by changing from full correlation matrix computation to PSD difference-based transfer function estimation. This parameter transformation reduces the computational complexity from O(N^2) for N microphones to a much more efficient calculation that scales linearly with the number of microphone pairs, enabling accurate noise suppression in large arrays with limited computational resources.
Data Source
AI summary
Noise is suppressed from a microphone array by estimating a noise field isotropy. In some examples audio is received from a plurality of microphones. A power spectral density of a beamformer output is determined and a power spectral density of microphone noise differences is determined. A noise power spectral density is determined using a transfer function and the noise power spectral density is applied to the beamformer output power spectral density to produce a power spectral density output of the received audio with reduced noise.