The study of sleep phases is helpful to evaluate the sleep state and quality. In order to explore the relationship between heart rate variability (HRV) and sleep phases, detrended fluctuation analysis (DFA) and autoregressive (AR) model spectrum estimation were used to analyze R-R interval (RRI) sequence of 45 healthy subjects from Sleep Heart Rate and Stroke Volume Data Bank. The results show that, in different sleep phases, long-range correlation was detected in RRI sequence. From rapid eye movement to light sleep and deep sleep, scaling exponent alpha1 and alpha2 were decreasing, while spectral parameters LFn. u., HFn. u. and LF/HF showed increasing parasympathetic activity. Scaling exponents had more significant differences than spectral parameters in distinguishing different sleep phases. Thus, DFA will be a better analysis method to quantify HRV in different sleep stages.