IF Estimation of Overlapped Multicomponent Signals Based on Viterbi Algorithm

Viterbi algorithm (VA) on time frequency (TF) distribution is a highly performed instantaneous frequency (IF) estimator. However, inaccurate IFs may be tracked due to switch problem in VA when signal components are overlapped on the TF plane. In order to address the problem, this paper first assumes the IF linearity in the overlapped TF regions should not change much, then, a new penalty function describing the variation of IF linearity based on the linear least square fitting technique is developed, and finally, a novel algorithm composed of two IF estimates is introduced. In the first coarse IF estimation, original VA is applied to determine the TF overlapped regions. In the second fine IF estimation, a modified VA employing the new penalty function is applied in the overlapped regions, while the original VA still functions in the non-overlapped regions. Simulations indicate the proposed algorithm can effectively suppress the switch problem and thus can achieve accuracy improvement especially for non-monotonous IF curves compared to other VA-based estimators.