The analysis of heart rate variability (HRV) has become a tool for noninvasively detecting the cardiovascular modulation of autonomic nervous system. Traditional analysis in frequency domain mainly includes calculating the power and detecting the peak frequency of each physiological frequency band. Whether employing the classical method or AR model to estimate the spectrum, the approximate stationarity of HRV is presupposed. Only in short term analysis can data meet this condition, while in long term the nonstationarity of HRV notably appears. A dynamic analysis method based on wavelet transform was proposed in this paper, which not only can obtain the traditional indices in frequency domain, but can compute their dynamic values varying with time, called short-time power and short-time LF/HF ratio. The latter can dynamically evaluate the activity of autonomic nervous. Finally the method was applied to trace the balance of autonomic nervous in Atropin drag experiment.