A non-line-of-sight identification algorithm in automatic microphone localization for experimental acoustic modal analysis

In the past few years an increasing interest in the minimization of interior noise levels has been observed in the automotive industry. Within this context, experimental Acoustic Modal Analysis (AMA) tests play a crucial role in the optimisation process of the acoustic parameters of car cavities. A first challenging task of such tests is the creation of an accurate geometrical wireframe model representing the 3D positions of the microphones instrumented inside the car cabin. In this paper, an automatic microphone localization procedure is presented and validated with real experimental data. The technique is based on multilateration concepts: four or more acoustic sources (anchors) are utilized for measuring the time-of-arrival (TOA) between sources and microphones (targets), so that the targets can be localized in 3D space. However, typical obstructions inside a car cabin lead to non-line-of-sight (NLOS) links between the anchor and the target. As a consequence, range estimates may have an erroneous positive bias. A robust technique is proposed, which identifies and prunes the NLOS measurements, so that the microphones can be localized using the LOS distances only. An experimental case study on a fully trimmed sedan car will be shown in order to demonstrate the effectiveness of the algorithm.