Automatic sleep stage classification based on ECG and EEG features for day time short nap evaluation

In this study, the Electrocardiogram (ECG) and Electroencephalogram (EEG) data recorded during day time short nap were analyzed. The ultimate purpose is to find out effective ECG features combined with usual EEG features for sleep stage determination during day time nap. Firstly, the ECG data was pre-processed in order to eliminate artifacts. After preprocessing, the second-order derivative of the ECG signal was calculated and clustered into two classes by K-means method. The peak positions of R wave were detected. Secondly, the Heart Rate Variability (HRV) was calculated according to the RR intervals (RRIs). Features of HRV of ECG were extracted in time-domain and frequency-domain. The redundant features were removed by the rough set method. Finally, the extracted features from the HRV of ECG were combined with the usual EEG features for sleep stage determination. The sleep stages including stage awake, stage 1 and stage 2 were distinguished by using Support Vector Machine (SVM). The obtained result indicated that the extracted ECG features improved the sleep stage classification accuracy.