1：Experimental Design and Method Selection:

The study involved designing a portable hardware platform for PPG signal acquisition, defining and extracting 125 morphological features from PPG and its derivatives, using six feature selection methods (Spearman correlation, ReliefF, information gain, chi-square, mRMR, Gini index) to evaluate features, and establishing linear and nonlinear classification models (LDA, cubic SVM, weighted KNN, logistic regression) for hypertension risk stratification.

2：Sample Selection and Data Sources:

Data were collected from 219 subjects recruited from Guilin People's Hospital, with exclusions for certain diseases and medications, resulting in a final dataset of 124 subjects categorized into normotensive, prehypertensive, and hypertensive groups based on BP criteria.

3：List of Experimental Equipment and Materials:

A customized PPG probe with an MSP430FG4618 MCU, infrared transmission mode at 905-nm wavelength, 1 kHz sampling rate, 12-bit ADC,

4：5–12 Hz bandpass filter, and an Omron 7201 electronic sphygmomanometer for BP measurement. Experimental Procedures and Operational Workflow:

Participants rested for over 30 minutes, then PPG signals were collected from the left index finger and BP from the right forearm simultaneously within 3 minutes using the custom probe and Omron device; three

5：1-s PPG records per subject were saved, and physiological data were recorded via a mobile app. Data Analysis Methods:

Data pre-processing included signal quality evaluation using skewness method, bandpass filtering, mathematical derivatives for VPG and APG, feature extraction, z-score normalization, feature selection, and classification with 10-fold cross-validation using MATLAB R2017a.