Chiesa L., Nijman E.

Inverse sound source reconstruction seeks, given a transmission matrix, to reconstruct the volume velocity distribution on the source surface using sound pressure data. In practice this discrete inversion problem turns out to be ill-conditioned and needs regularisation in order to produce stable results. The core issue remains the determination of the appropriate degree of regularisation, such as to guarantee a stable solution with a minimal loss in spatial resolution for the reconstructed source. In this paper a new optimisation criterion is presented for the Tikhonov regularisation parameter. This new method uses errors expressed in decibels instead of the conventional squared pressure differences to estimate the accuracy of the reconstructed field. Such procedure allows to make a better use of the information carried by sensors with a relatively low signal. Laboratory experiments have been carried out and the performance of the new criterion has been compared with those of the conventional methods using L-curve and cross validation techniques.