A new least-squares based method for selection of spectral peaks from the spectrum estimated by sparse representation

Abstract

In this paper, a new method for selection of spectral peaks is proposed, when the spectrum is estimated based on sparse representation. The proposed method fits the spectral peaks to the available data using least squares fitting, and then computes the remaining signal. If the remaining signal contains noise only, then all the spectral peaks are detected. Computer simulations verify that the proposed method is comparable to the case where the number of spectral peaks is known.