Table 2.

Common characteristics of PSG in arousal detection with machine learning methods.

Channel Name	The Discussed Features and the Related References
EEG	Spectral energies in the delta, theta, alpha, beta, and gamma bands [49]
	Approximate entropy (ApEn) [50]
	Power spectrum density [51]
	Wavelet packet decomposition (WPD) [52]
	Hjorth parameters (including Hjorth activity, mobility, and complexity) [53]
	Wavelet transform [54]
	Frequency and amplitude [45]
EOG/chin EMG	Spectral energies in the delta, theta, alpha, beta, and gamma bands [49]
	Form factor, standard deviation, skewness, kurtosis, and relative energies [55]
	Submental, amplitude [45]
CHEST/ABDOMINAL/AIRFLOW	Breath rate, width, amplitude, inspiratory, slope, inter-breath intervals [56]
	Coefficient of variation of the signal envelope [57]
	Form factor, standard deviation, skewness, kurtosis, and relative energies in two regions [55]
	Respiratory disturbance variable (RDV) [57]
	Correlation between abdomen and thorax signals [58]
SaO2	Rolling mean [59]
	Hypoxic burden, proportion, standard deviation, skewness, kurtosis [60]
	Statistical features [59]
ECG	Heart rate, inter-beat intervals, and R-wave amplitude time-series [36]
	Rolling variance [59]
	QRS [61]
	Heart rate variability (HRV) signals [62]