Sleep apnea syndrome (SAS) is monitored and examined clinically with polysomnography. However, it is expensive and complex to operate, which significantly affects the natural sleep of human. To evaluate the value of heart rate variability (HRV) in diagnosing SAS, we propose a new method for SAS classification based on fuzzy support vector machine (FSVM). Detrended Fluctuation Analysis (DFA) and Autoregressive (AR) model spectrum estimation are used to analyze R-R interval sequence of 38 healthy subjects and 28 SAS subjects during various sleep stages. Scaling exponents of age, gender and HRV at each sleep stage, as well as low/high frequency are selected as SAS characteristic parameters and FSVM is used to classify SAS. Results indicate that the proposed method can diagnose SAS effectively and the classification accuracy rate of SAS is 93.94%. Compared with current SAS diagnosis methods, this method is more simple and efficiently