正交匹配追踪反卷积声源识别方法

OMP-DAMAS波束形成声源识别方法能够显著缩减主瓣宽度,降低旁瓣水平,获得极高的分辨率和定位精度。基于数值仿真多个声源激励下的识别成像图和偏差值探究其结果随声源频率、迭代次数和信噪比等参数的变化规律。结果表明:OMP-DAMAS能够有效提高分辨率和定位精度,适用于中高频声源的识别并且对噪声具有较好的适应性。当声源频率大于2 300Hz,信噪比高于0dB时,OMP-DAMAS均能准确识别声源,获得清晰的成像结果。其重构的声源个数取决于迭代次数,在信噪比较高时可以通过设置合适的动态范围以避免旁瓣污染。上述结论对反卷积OMP-DAMAS波束形成技术的运用具有指导意义。进一步,基于多个扬声器的声源识别试验验证了该方法的有效性。

OMP-DAMAS beamforming acoustic source identification

Orthogonal matching pursuit applied to the deconvolution approach for the mapping of acoustic sources inverse problem (OMP-DAMAS) is characterized with high resolution and high location accuracy by reducing mainlobe and eliminating sidelobes.The change rules of the acoustic source identification results with signal frequency, iteration number and signal to noiseratio (SNR) were explored based on numerical simulated incoherent sources identification imaging maps and deviation value. The results are listed as following. OMP-DAMAS can improve resolution and location accuracy effectively. Excellent performance and robustness are guaranteed for medium and high frequency source. Acoustic sources can be identified with clear map when the algorithm is applied to sources with frequency higher than 2 300 Hz and SNR higher than 0 dB. The number of reconstructed sources depends on iteration number, while sidelobes can be eliminated by setting dynamic range properly. These conclusions have guiding significance for the application of the OMP-DAMAS. Furthermore, the feasibility of the method has been validated through a physical experiment on acoustic source identification of several loudspeakers.