Electrophysiological signal segmentation for EEG frequency bands and heart rate variability analysis

Biological signals like Electroencephalography (EEG) and Electrocardiography (ECG) provide very useful information related to brain and heart condition. Spectral analysis of EEG frequency bands and analysis of Heart Rate Variability (HRV) of ECG signal are largely used as a biomarker in studying various pathological conditions. However, the studies on EEG and HRV frequency bands analysis have shown inconsistency in the size of segment length used to extract the features. In the power spectral analysis, the trade-off between time and frequency resolution remain unsolved issue. To increase a frequency resolution and reduced a spectral leakage, a long segment of input data is necessary. However, for a stationary consideration a short segment data is needed. Therefore, in this study we examine the effect of segmentation time window and overlapping in spectral analysis of EEG frequency bands and HRV. The EEG and ECG signals used in this study are collected from the healthy and sleep apnea patients. Five features (delta, theta, alpha, sigma and beta) are extracted from EEG frequency bands and three (LFnu,HFnu, LF/HF) from the HRV by using spectral analysis at various window size and overlapping percentage. The results show the optimum window size for EEG is achieved at 10 sec and 120 sec for HRV analysis.