Table 4 Summary of handcrafted feature extraction methods for potential BP monitoring.

Waveform features

The following features are extracted from the waveform of a signal or its 1st/2nd derivative:

• Peak/valley/key point amplitude

• Amplitude difference (height) between two key points

• Time duration (width) between two key points

• The area under the waveform during a certain period

• Angle/slope

PPG

^{111,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174,175,176}

ECG

^{173,174,175,177,178}

US

^179,180,181

TAG

^182,183

IPG

⁸

PTT/PAT/PWV-based features

The time interval between two reference points in two signals of the same or different types. PWV-based features take both transit time and distance into consideration. Reference points include but are not limited to:

• ECG R-peak

• PPG peak/foot/maximum 1st derivative point

• IPG B-point

• TAG systolic/diastolic peaks, etc.

TAG + TAG

²⁷

PPG + ECG

^{105,164,173,175,177,184,185,186,187}

PPG + TAG

¹⁸⁸

ECG + TAG

^17,182

ECG + IPG

⁷

IPG + IPG

⁸

PPG + IPG

^49,97

Frequency domain features

Signals are transformed to the frequency domain to extract spectral features.

PPG

¹⁷⁰

PIR-based features

The ratio between signal systolic and diastolic amplitude is an indicator for BP.

PPG

^{175,177,189,190,191}

TAG

¹⁸²

PAR

PPG acceleration ratio

PPG

¹⁹²

Statistical features

Statistics of raw signal sequences or extracted parameters, e.g., the standard deviation of RR intervals

PPG

¹⁹³

ECG

¹⁹⁴