Respiratory rate derived from smartphone-camera-acquired pulse photoplethysmographic signals

Abstract

A method for deriving respiratory rate from smartphone-camera-acquired pulse photoplethysmographic (SCPPG) signal is presented. Our method exploits respiratory information by examining the pulse wave velocity and dispersion from the SCPPG waveform and we term these indices as the pulse width variability (PWV). A method to combine information from several derived respiration signals is also presented and it is used to combine PWV information with other methods such as pulse amplitude variability (PAV), pulse rate variability (PRV), and respiration-induced amplitude and frequency modulations (AM and FM) in SCPPG signals. Evaluation is performed on a database containing SCPPG signals recorded from 30 subjects during controlled respiration experiments at rates from 0.2 to 0.6 Hz with an increment of 0.1 Hz, using three different devices: iPhone 4S, iPod 5, and HTC One M8. Results suggest that spontaneous respiratory rates (0.2-0.4 Hz) can be estimated from SCPPG signals by the PWV- and PRV-based methods with low relative error (median of order 0.5% and interquartile range of order 2.5%). The accuracy can be improved by combining PWV and PRV with other methods such as PAV, AM and/or FM methods. Combination of these methods yielded low relative error for normal respiratory rates, and maintained good performance at higher rates (0.5-0.6 Hz) when using the iPhone 4S or iPod 5 devices.