In order to study the nonstationary heart rate variability (HRV) signals under gravitational stress, a time-variant autoregressive spectral estimation method has been implemented, which made it possible to study dynamically the process of cardiac autonomic regulation during various gravitational stresses. The HRV signals obtained from 8 subjects during seated-LBNP testing before and after a six-month aerobic training were analyzed by the implemented algorithm. The obtained time-variant spectra might clearly reveal the time course of changes of decrease in HFn and increase in LF/HF, reflecting the time course of cardiac vagal withdrawal and indices with time during seated-LBNP were increased in most of the subjects. In brief, our work shows that the method of time-variant AR spectrum analysis for non-stationary HRV signals makes it possible to elucidate the dynamics of cardiac autonomic regulatory activity under gravitational stress and to detect the subtle changes in HRV after aerobic training.