Sparsity-based frequency-hopping spectrum estimation with missing samples

Abstract

In this paper, we address the problem of spectrum estimation of frequency-hopping (FH) signals in the presence of random missing samples. The signals are analyzed within the bilinear time-frequency representation framework, where a time-frequency kernel is designed based on inherent FH signal structures. The designed kernel permits effective suppression of cross-terms and artifacts due to missing samples while preserving the FH signal auto-terms. The kernelled results are represented in the instantaneous autocorrelation function domain, which are then processed using sparse reconstruction methods for high-resolution estimation of the FH signal time-frequency spectrum. The proposed method achieves accurate FH signal spectrum estimation even when a large proportion of data samples is missing. Simulation results verify the effectiveness of the proposed method and its superiority over existing techniques.