Search

To date, the applications of photoplethysmograms in the estimation of cuff-less blood pressure, arteriosclerosis, and so on have been studied. Photoplethysmogram waveform features have often been used to estimate the target volumes. For these estimations, it is necessary to acquire photoplethysmogram waveforms and changes in their derivative waveform details, for which the photoplethysmograms measured at comparatively high sampling rates have been used. The performance of smartphone cameras and wearable photoplethysmogram sensors has improved; with regard to mobile health technology, photoplethysmograms measured at lower sampling rates offer considerable advantages. These include lower computational resources, compression of accumulated data, and lower sensor power consumption. However, compared to photoplethysmograms measured at a high sampling rate, photoplethysmogram measurement at a low sampling rate will result in waveform signal degradation. This paper investigates the possibility of using photoplethysmograms measured at a low sampling rate. To this end, we statistically compared photoplethysmogram waveform features obtained from 63 male subjects free of circulatory diseases, at a sampling rate of 240 Hz, with waveform features obtained at low sampling rates (120, 60, 30, 20, and 10 Hz) through downsampling, and evaluated possible commercial use.

A pulse waveform profiling algorithm is introduced for processing real-time photoplethysmography (PPG) data in mobile applications. The real PPG waveform is discriminated from artifacts based on rules of probability distribution and physiological significance. The swift background is removed through a two-step method to minimize its effect on the final waveform analysis: interpolating the midpoint of the amplitude of the PPG AC part and then interpolating the minimum of each pulse period. The pulse waveform features are extracted from the stratified PPG wave data. The implementation of the proposed algorithm in a real-time mobile application demonstrated its high efficiency and robustness with a sensitivity of around 96% and a positive specificity of 100%. The potential applications of these extracted features are discussed in terms of long-term trends in physiology such as circadian and arrhythmia real-time monitoring.