A highly efficient compressed sensing algorithm for acoustic imaging in low signal-to-noise ratio environments. (November 2018)

• Record Type: Journal Article
• Authors: Ning, Fangli; Pan, Feng; Zhang, Chao; Liu, Yong; Li, Xiaofan; Wei, Juan
• Is Part Of: Mechanical systems and signal processing. Volume 112(2018)
• Page Start: 113
• Page End: 128
• Publication Date: 2018-11
• Subjects: Compressed sensing -- Microphone array -- Acoustic imaging -- Singular value decomposition -- Highly efficient; 00-01 -- 99-00; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2018.04.028
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: We propose an OMP-SVD method for acoustic imaging, which combines the orthogonal matching pursuit (OMP) algorithm of compressed sensing and singular value decomposition (SVD). Compared with the conventional beamformer (CBF) method, the OMP method and the l1 -SVD method, our proposed method is more robust in low signal-to-noise ratio (SNR) and low frequency environments. Besides, our proposed method is highly computationally efficient. Our OMP-SVD method can obtain good results when there is basis mismatch. We also find that our proposed method performs better than other methods when the target sparsity is overestimated. The conducted experiment results prove that the OMP-SVD method can obtain satisfying results in low SNR and low frequency environments, even if the target sparsity KT is overestimated. Abstract: We study the acoustic imaging in low signal-to-noise ratio (SNR) environments with compressed sensing (CS) and microphone arrays. In this work, we propose an OMP-SVD method which combines the orthogonal matching pursuit (OMP) method of CS and the singular value decomposition (SVD). The performance of the proposed OMP-SVD method is compared with the CBF method, the OMP method and the l 1 -SVD method. In terms of the CPU time, the proposed method is highly efficient like the CBF method and the OMP method, and much more efficient than the l 1 -SVD method. In terms of the accuracy of the source maps, the OMP-SVD method can locate the sources exactly for the SNR as low as −10 dB and the frequency as low as 2000 Hz, while the other three different methods can only locate the sources when the SNR is greater than or equal to 5 dB. In addition, we find that the proposed method can obtain good performance when the target sparsity K T is overestimated and there is basis mismatch. Finally, a gas leakage experiment was conducted to verify the performance of the OMP-SVD method in practical application. The results show that the OMP-SVD method is robust in low SNR environments.