ECG frequency analysis is complicated by the fact that ECG signals are non-stationary, that is, their activity patterns change slowly or intermittently as a result of variations in a number of physiological and physical influences. Previous applications of spectral methods to the ECG analysis have identified power spectrum of the intervals between heart beats, characteristic frequencies of which belong to low frequency ranges. Here we introduce a new approach, based on the recently developed SBF algorithm of numerical Fourier transform spectroscopy, which can evaluate frequency transfer functions from electrical transients induced by the heart activity. Using this approach, we found that ECG components, induced by a heart contraction are consistent with the transient behavior of a classical underdamped harmonic oscillator. Identification of this model using the Bode diagram, allows us to characterize heart contractions by the novel parameters: natural frequency and damping ratio of heart contraction. Characteristic natural frequency of a healthy heart is in the frequency range between 13 and 20 Hz.