Table 2.

Performance of the CNN and the CNN-LSTM models for the three-level drowsiness classifications for different combinations of input signals.

Model Name	Input Signal	Accuracy (%)	Sensitivity (%)	Specificity (%)	Precision (%)
CNN	HRV	55	58	59	56
	HRV-PSD	63	61	61	59
	RR	66	62	61	65
	HRV + HRV-PSD	65	63	60	62
	HRV + RR	72	67	71	70
	HRV-PSD + RR	68	64	65	65
	All inputs	84	80	88	80
CNN-LSTM	HRV	72	65	71	61
	HRV-PSD	73	65	72	64
	RR	81	76	83	74
	HRV + HRV-PSD	78	71	81	70
	HRV + RR	82	77	84	80
	HRV-PSD + RR	77	71	79	70
	All inputs	91	87	92	87

HRV: heart rate variability, HRV-PSD: power spectral density of HRV, RR: respiration rate.