Sound Source Identification Using Sparse Prior Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sound source identification methods using microphone arrays suffer from low spatial resolution and low computational efficiency, particularly in accurately locating and intensifying sound sources.
Innovation Solution
A sound source identification method based on array measurement and sparse prior information, which involves building a sound source identification problem model, extracting peaks from a basic sound image, determining the number and distribution of support sets, and using proximal gradient iteration to generate a high-resolution sound image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional beamforming method based on delay-sum is used, then calculation speed is fast and robustness is good, but spatial resolution is low and accuracy is poor
Solution Approach 1:
The patent applies preliminary action by using conventional beamforming to generate a basic sound image first, then extracting peak positions from this preliminary result to initialize the support set for subsequent high-resolution reconstruction. This preliminary step provides a good starting point that accelerates the convergence of the iterative optimization algorithm, thus improving final accuracy without sacrificing calculation speed.
Solution Approach 2:
The patent segments the sound source identification process into two distinct stages: (1) conventional beamforming to obtain basic spatial information and peak positions, and (2) high-resolution reconstruction using iterative optimization with support set constraints. This segmentation allows each stage to focus on its strength, achieving both speed and accuracy.
2Measurement precision
If sound field reconstruction method based on acoustic inverse problem with sparse constraint is used, then spatial resolution is improved and super-resolution is achieved, but computational complexity increases and calculation efficiency decreases
Solution Approach 1:
The patent applies local quality by introducing a support set that locally constrains the solution space to only those grid points identified as potential sound source locations from the basic sound image. This local constraint reduces the computational burden of the iterative optimization algorithm by focusing calculations only on relevant regions, thereby improving computational efficiency while maintaining high spatial resolution.
Solution Approach 2:
The patent uses preliminary action by performing conventional beamforming first to identify peak positions, then using these positions to define the support set for the subsequent high-resolution reconstruction. This preliminary identification step significantly reduces the search space and computational complexity of the iterative optimization process.
3Object-affected harmful factors
If adaptive beamforming method based on statistical characteristics is used, then interference suppression is improved, but spatial resolution is limited by Rayleigh criterion and not suitable for multiple sound sources
Solution Approach 1:
The patent applies parameter changes by transitioning from the Rayleigh limit inherent in conventional beamforming to super-resolution capabilities through iterative optimization with support set constraints. By changing the mathematical approach from direct beamforming to constrained optimization, the system achieves spatial resolution beyond the Rayleigh criterion while maintaining interference suppression capabilities.
Data Source
AI summary
The present disclosure provides sound source identification based on array measurement and sparse prior information, which uses a generalized minimax concave penalty function to apply a sparse constraint on a sound source reconstruction error, determines position information of a potential sound source according to a basic sound image generated by a conventional beamforming method, and generates an initial solution vector including a sound source position prior information in a specific way. Compared with using a random initial value and a zero initial value, computational efficiency of the disclosed sound field reconstruction method is improved. Further, an iterative convergence speed is improved by using a proximal gradient acceleration and an adaptive step size backtracking strategy, an optimal step size that can meet convergence is adaptively selected, and sound field reconstruction can be completed more quickly. The present disclosure realizes high-resolution and high-efficiency sound source identification by using a microphone array.