Heart rate variability (HRV) has the characteristics of non stationary and transient fluctuation. In this paper, we propose using the autoregressive conditional heteroskedasticity (ARCH) to analyze such fluctuations. Analytical methods include using auto-regressive moving average model to eliminate sequence trends and correlativity, using the F-test method to determine whether there is ARCH effect in residual sequence, while giving an acceptance probability of ARCH effect. The fluctuations of HRV in congestive heart failure (CHF) patients and normal people are analyzed in this paper. The large difference in statistics value of acceptance probability between CHF patients and normal people indicates that the method can well distinguish between different groups. It provides a new method for the study of ECG information.