Reconstruction of Acoustic Signals According to Incomplete Data

摘要

Problems of reconstructing acoustic signals from unevenly spaced samples (sparse signal) and signals distorted by losses of adjacent samples (solid gaps) are considered. To solve these problems original algorithms are proposed: the Interpolation Method of Sequential Computation of the Fourier spectrum and the method of amplitude iterations. The effectiveness of these methods is compared with the method of projections onto convex sets and its modification, implemented using an evolutionary time-frequency transformation based on the basic functions of Slepian. In the case of continuous gaps, comparisons are made with a one-dimensional modification of the method, which is related to neural networks image inpainting. Evaluation of the effectiveness of the proposed methods according to objective criteria indicates their suitability for practical use.