Time domain localization technique with sparsity constraint for imaging acoustic sources. (15th September 2017)

• Record Type: Journal Article
• Authors: Padois, Thomas; Doutres, Olivier; Sgard, Franck; Berry, Alain
• Is Part Of: Mechanical systems and signal processing. Volume 94(2017)
• Page Start: 85
• Page End: 93
• Publication Date: 2017-09-15
• Subjects: Localization -- Inverse method -- Sparsity constraint -- Workplace; Structural dynamics -- Periodicals; Vibration -- Periodicals; Constructions -- Dynamique -- Périodiques; Vibration -- Périodiques; Structural dynamics; Vibration; Periodicals; 621
• DOI: 10.1016/j.ymssp.2017.02.035
• Languages: English
• ISSNs: 0888-3270

Abstract

Highlights: A source localization technique based on inverse problem with sparsity constraint is proposed. Numerical results show that the proposed technique allows for high resolution imaging. Experimental validation shows that the technique is able to localize the main source and its reflections. Abstract: This paper addresses source localization technique in time domain for broadband acoustic sources. The objective is to accurately and quickly detect the position and amplitude of noise sources in workplaces in order to propose adequate noise control options and prevent workers hearing loss or safety risk. First, the generalized cross correlation associated with a spherical microphone array is used to generate an initial noise source map. Then a linear inverse problem is defined to improve this initial map. Commonly, the linear inverse problem is solved with an l 2 -regularization. In this study, two sparsity constraints are used to solve the inverse problem, the orthogonal matching pursuit and the truncated Newton interior-point method. Synthetic data are used to highlight the performances of the technique. High resolution imaging is achieved for various acoustic sources configurations. Moreover, the amplitudes of the acoustic sources are correctly estimated. A comparison of computation times shows that the technique is compatible with quasi real-time generation of noise source maps. Finally, the technique is tested with real data.