Robust sparse reconstruction of signals with gapped missing samples from multi-sensor recordings

Abstract

Reconstructing missing samples in multi-sensor recordings is important in many applications including blind source separation, signal detection and direction of arrival estimation problem. This study proposes a method of reconstructing missing samples from multi-sensor recordings of non-stationary amplitude-modulation frequency-modulated signals using a two step procedure. In the first step the instantaneous frequency of a given multi-sensor signal is estimated using a robust and efficient method. Then the instantaneous frequency information is exploited to jointly recover the instantaneous amplitudes of multi-sensor signals. For recovering instantaneous amplitudes, the Fourier series approximation of the instantaneous amplitude is exploited. Experimental results indicate that the proposed method achieves better performance than the existing method including time-frequency filtering method, gradient descent and orthogonal matching pursuit algorithms.