TABLE 2.

The machine learning models and traditional polynomial models.

	References	Parameters	Detail methods	Number of subjects	MAE (mmHg)		Standard deviation (mmHg)		Dataset	Subject description
					DBP	SBP	DBP	SBP
Machine learning related models	Chowdhury et al. (84)	PPG features	Gaussian process regression	219	3.02	1.74	5.54	9.29	Public data (87)	657 PPG signal samples from 219 subjects.
	Aguirre et al. (86)	PPG wave	Deep learning with attention mechanism	1100	6.57	14.39	8.43	17.87	MIMIC-III	10,696 segments corresponding to 1131 subjects.
	Xing et al. (83)	PPG features, BMI	Random forest algorithm	1249	2.3 (Young, Fitting error)	2.1 (Young, Fitting error)	9.5 (Young)	13.6 (Young)	Specific data	A total of 2358 measurements were recorded, including young, old populations. Normal, pre-hypertension and stage I, II, III hypertension.
	Watanabe et al. (85)	PPG features	Specific algorithm based on second derivative of photeplethysmogram wave.	887	NA	NA	NA	NA	Specific data	A total of 887 participants were enrolled. Various feature parameters of pulse wave participants at rest an under exercise, mental stress are collected.
	Zhang et al. (30)	PTT (HRV, ECG, PPG, other PPG features)	LR: linear regression SVR: support vector regression RF: random forest regression Adaboost: adaptive boosting	3337	5.35	10.03	4.5	7.96	MIMIC I and VitalDB	Hybrid dataset (including 3,337 subjects) combining MIMIC and VitalDB databases.
Traditional polynomial models	Ghosh et al. (79)	PTT (R peak of ECG and peak of PPG)	$BP=\frac{a}{PTT}+b$	14	6.64 (Recumbent)	4.6 (Recumbent)	5.2 (Recumbent)	9.6 (Recumbent)	Specific data	14 subjects performed activities including: recumbent, seated, standing, walking, cycling, need calibration.
	Huynh et al. (80)	PTT	$DBP=DB{P}_0+\rho \frac{D^2}{PT{T}^2}1\mathrm{n}\left[1+K(Z_{\max 0}-Z_{\min })\right]$(A.2a) $SBP=DB{P}_0+\rho \frac{D^2}{PT{T}^2}1\mathrm{n}\left[1+K(Z_{\max 0}-Z_{\min })\right]$(A.2b)	15	5.02 ± 0.73 (RMSE)	8.47 ± 0.91 (RMSE)	NA	NA	Specific data	15 young, healthy human subjects leveraging handgrip exercises.
	Esmaili et al. (81)	PTT	$\text{BP}=a_0+\sqrt{a_2+a_2\frac{1}{{\mathrm{PTT}}^2}}$	32	3.97	6.22	5.15	9.44	Specific data	32 healthy subjects in the age range of 21–50 years performed physical exercise.
	Lin et al. (82)	PTT (PPG features)	Linear regression and four previously reported models (88–91).	22	3.16 (DS)	3.19 (DS)	5.04 (DS)	7.8 (DS)	Specific data	22 subjects when they performed mental arithmetic stress and Valsalva’s manoeuvre tasks that could induce BP fluctuations.