Improved identification of sound sources at low frequencies using the two-step steepest descent method based on the equivalent source method

Abstract

Source identification is challenging when the underdetermined systems consist of relatively fewer measurement data than the sources to reconstruct. Wideband acoustic holography (WBH) is a sparse regularization-based source identification method with satisfactory reconstruction accuracy. In the previous work, it is found that the initial source strength distribution significantly affects the results, so it is challenging to identify low-frequency coherent sources using WBH alone. In this work, the improved WBH is proposed, motivated to refine its deficiencies at low-frequency reconstruction. The initial source strength of WBH is modified by introducing the results of the fastest descent method. Combined with the a priori assumption of the source strength, label the equivalent source of the actual source and adjust the threshold filtering conditions to ensure that these equivalent source strengths are not involved in the filtering process during the iteration. The present work shows that the reconstruction accuracy of the modified WBH is similar to that of the WBH at mid-high frequencies. Meanwhile, it can bear better reconstruction and localization accuracy at low frequencies without degrading the reconstruction performance at the mid to high-frequency sound field.