A Robust Approach to the Order Detection for the Damped Sinusoids Based on the Shift-Invariance Property

In this paper, the problem of robust order detection for the damped sinusoids in impulsive noise environments is addressed. First of all, a series of candidate models are assumed. It is a challenging issue to extract the potential signal subspaces for these candidate models in a computationally efficient way. To alleviate this difficulty, the successive robust low-rank decomposition (SRLRD) procedure is devised. Correspondingly, the values of the subspace-based automatic model order selection (SAMOS) criterion are calculated. As a result, the signal order estimate is determined by the candidate model with the minimum criterion value. The consistency and superiority of the proposed estimator is validated by the simulation results in comparison with the other existing schemes.