弹性网正则化广义逆波束形成算法改进

广义逆波束形成凭借其空间分辨率高,抑制旁瓣能力强等优势得到广泛关注。为了提升一般广义逆波束形成的声源识别性能,基于弹性网正则化波束形成既能保证声源识别结果的稳健性又能体现声源信号的稀疏性。然而,在测量声源信号的过程中所产生的非相干噪声对声源结果产生不可避免的误差,为抑制测量过程的干扰噪声,结合对角降噪和特征值改进法重构波束形成正则化参数,提出了一种改进弹性网正则化的广义逆波束形成,以重构正则化参数区别干扰噪声和目标声源。进行了数值仿真和实验验证,结果表明该算法在中高频时主瓣宽度误差低于10 dB,且比弹性网正则化波束形成具有更高的空间分辨率以及稳健性,衰减旁瓣能力强。

Generalized inverse beamforming with improved elastic net regularization

The generalized inverse beamforming has attracted wide alttention due to its high spatial resolution and strong sidelobesuppression eapabiliies. To improve the sound souree identifieation performance of GIB, a regularized beamforming method via elasticnet is proposed, which ean ensure the robustness and sparsity of the sound souree identification results, However, the ineoherent noisegenerated in the proeess of measuring the sound source signal may produces unavoidable errors. To suppress the interference noise in themeasurement process, a generalized inverse beamforming method with improved elastie net regularization is proposed by combiningdiagonal denoising with improved eigenvalue method to reconstruct the regularization parameters of beamforming, which could distinguishthe interference noise from the target sound source. Numerical simulation and experimental results both prove that the main lobe widtherror of the proposed algorithm is less than 10 dB at high frequencies, and it has higher spatial resolution and robustness, and strongsidelobe attemuation ability than elastie net regularized heamforming as well.