Identification of acoustic moving sources in the context of a road vehicle at pass-by: A numerical comparison

Abstract

The problem of the identification of acoustic sources has been widely treated in static or quasi-static source contexts, more recently for moving sources, in a wide range of applications. Different methods have been developed in this endeavor. In the context of identification of sources from a passing-by vehicle, the beamforming method is a reference that has well-known limitations: a poor resolution at low frequencies causes difficulties to discriminate close sources or with too different levels, and the convolution of the point spread function with the source-signal makes it difficult to evaluate the level precisely. Deconvolution methods are used to overcome those limitations. Mostly developed in static contexts at the beginning, extensions to moving sources have been proposed in the transportation field, mainly in air transportation and underwater acoustics. In the present investigation, these methods are numerically tested with parameters fitted to the road vehicle context. Their performance is assessed at different speeds, with varying additional noise levels. Beamforming—as the reference method—and several deconvolution methods (DAMAS, CLEAN, NNLS) extended to moving sources are compared using performance including: location of the maximum, source gravity center, maximum level and extended level. Illustrations involving an academic pendulum setup are given.