This paper develops a recursive algorithm for tracking resonance frequency shifts of linear timevarying systems corrupted by additive white Gaussian noises, in real time. So far, automatic resonance tracking is limited to non-model-based designs that rely solely on the phase difference between a selected input and output of the system. Instead, we propose a transformation of the system into a complex-valued representation, which allows to abstract the resonance shifts as an exogenous disturbance acting on the excitation frequency, perturbing it from the natural frequency of the plant. The resonance tracking problem is then split in two tasks; identifying the frequency disturbance and solving an optimization problem to determine the excitation frequency. The design of the resonance tracker is therefore simplified, due to the application of well-established techniques. We discuss the stability of the proposed scheme, even in cases that seriously challenge the current phasebased approaches, like non-monotonic phase differences and multiple-input multiple-output systems. Numerical simulations further demonstrate the performance of the resonance tracking scheme.Keywords resonance • frequency tracking • time-varying systems • complex variables • closed-loop identification.