Modified GEV Beamformer for Noise Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Smart speakers and voice-controlled devices face challenges in efficiently isolating target audio from noise in noisy environments, especially when the target audio can come from any direction relative to the microphones, and traditional beamforming solutions are processing intensive and unsuitable for diverse geometries and low-power devices.
Innovation Solution
The use of a modified generalized eigenvector (GEV) system that determines the Relative Transfer Function (RTF) of the target audio in real-time without knowing the microphone array geometry, employing a spatial filtering process like minimum variance distortionless response (MVDR) beamforming to enhance the target audio, which is computationally efficient and scalable for large microphone arrays.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If traditional beamforming solutions are used to isolate target audio from noise, then noise suppression capability is improved, but processing complexity and computational load increase significantly
Solution Approach 1:
The patent changes the fundamental parameters of the beamforming approach by using generalized eigenvector decomposition instead of traditional methods. This transforms the problem from optimizing weight vectors based on signal models to directly computing eigenvectors of covariance matrices, fundamentally changing the mathematical parameters and operations involved, thereby reducing computational complexity while maintaining noise suppression capability
Solution Approach 2:
The patent substitutes traditional beamforming mechanical/computational mechanisms with a statistical signal processing approach based on eigendecomposition. Instead of using model-based weight optimization or geometric phase shifting, the system replaces these mechanical-like operations with statistical analysis of signal covariance structures, achieving simpler computation with equivalent or superior performance
2Ease of operation
If traditional beamforming solutions are used to isolate target audio, then directional filtering is improved, but adaptability to diverse microphone geometries deteriorates
Solution Approach 1:
The patent creates a universal beamforming solution that works across diverse microphone geometries by using eigendecomposition of covariance matrices. The method is geometry-agnostic and can be applied to any microphone array configuration without requiring geometry-specific optimizations, thereby achieving multi-functionality and broad adaptability while maintaining directional filtering capability
Solution Approach 2:
Instead of adapting the beamforming algorithm to match specific microphone geometries (traditional approach), the patent inverts the problem by using a geometry-independent statistical approach. The system inverts the conventional wisdom by not requiring geometry information at all, using eigendecomposition to naturally adapt to any configuration, thereby simplifying the approach while enhancing versatility
3Measurement precision
If advanced beamforming algorithms are used to enhance target audio in real-time, then audio enhancement quality is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent changes the computational parameters from iterative optimization routines to direct eigendecomposition operations. By transforming the problem into an eigenvalue decomposition task, the system achieves closed-form solutions that are computationally more efficient than iterative methods, thereby improving computational efficiency while maintaining audio enhancement quality through accurate statistical modeling
Solution Approach 2:
The patent extracts the essential statistical characteristics of the audio signals through eigendecomposition, separating the signal subspace from the noise subspace. By extracting only the dominant eigenvectors corresponding to signal components and discarding others representing noise, the system achieves efficient computation focused on the most relevant signal features, improving both speed and quality
Data Source
AI summary
A real-time audio signal processing system includes an audio signal processor configured to process audio signals using a modified generalized eigenvalue (GEV) beamforming technique to generate an enhanced target audio output signal. The digital signal processor includes a sub-band decomposition circuitry configured to decompose the audio signal into sub-band frames in the frequency domain and a target activity detector configured to detect whether a target audio is present in the sub-band frames. Based on information related to the sub-band frames and the determination of whether the target audio is present in the sub-band frames, the digital signal processor is configured to use the modified GEV technique to estimate the relative transfer function (RTF) of the target audio source, and generate a filter based on the estimated RTF. The filter may then be applied to the audio signals to generate the enhanced audio output signal.