The minimal sampling frequency of the photoplethysmogram for accurate pulse rate variability parameters in healthy volunteers

Information Retrieval from Photoplethysmographic Sensors: A Comprehensive Comparison of Practical Interpolation and Breath-Extraction Techniques at Different Sampling Rates

“…Among time-domain parameters, Mean IBI was the only one showing little to no variation of the AEs among interpolation methods as the sampling rate decreased. This result finds confirmation in previous empirical work, such as [30], where interpolated and non-interpolated PPGs showed comparable accuracy in Mean IBI computation down to a sampling rate of 5 Hz. Further, it is supported by the theory since the discretization error is assumed as a zero-mean uniformly distributed white noise [34].…”

“…Given the notable amount of information that this signal can provide, many studies have focused on improving HRV parameters extraction from PPG (e.g., see [22][23][24]), and several others deal with the derivation of respiratory signal surrogates [21,[25][26][27]. However, only a few studies have systematically investigated the quality of HRV and respiratory features derived from low-sampling-rate PPG signals [28][29][30]. PPG has a much lower frequency content (generally, lying within the range 0.01-20 Hz [31,32]) than ECG (mostly 0-70 Hz in healthy subjects, with the need to reach at least 150-300 Hz for accurate characterization of the QRS complex [33]); thus, the identification of the fiduciary points in time is intrinsically affected by higher errors.…”

“…Specifically, HRV indices computed from PPG sampled at 100, 50, and 20 Hz, and processed with cubic-spline interpolation, were comparable to those extracted from the original PPG signal (collected at 200 Hz). In another study, Béres and colleagues [30] analyzed PPG signals recorded from participants at rest and concluded that sampling frequencies of at least 20 Hz and 50 Hz were required, respectively, with and without interpolation, to achieve accurate measures of time-domain and Poincaré plot indices. To reduce the computational cost required by spline interpolation, Baek et al [36] proposed the parabola approximation to interpolate PPG peaks and demonstrated its efficacy in PPG signals recorded at 20 Hz, compared to the spline approach.…”

Atrial Fibrillation Classification with Smart Wearables Using Short-Term Heart Rate Variability and Deep Convolutional Neural Networks

“…Firstly, most PPG based devices do not have a high sampling rate (~128 Hz), and vary from 60 Hz to 100 Hz based on the quality of the sensor and the battery levels of the device the sensor is embedded in. However, the minimum sampling frequency required is 50 Hz to derive reasonably accurate HRV and PRV parameters with a low margin of error from ECG and PPG signals, respectively [52,53]. Secondly, the computational overhead is reduced without a significant effect on the signal acquisition or processing aspects, which can extend the deployment of the proposed model in this work to resource-constrained wearable devices.…”

Pulse Rate Variability Analysis Using Remote Photoplethysmography Signals

“…As a result, 11 CPPGs and 11 RPPGs were generated for each subject. The sampling rate of RPPG was kept at 30; however, PRV analysis data with a sampling rate of 5 Hz can generally be used for calculating mean pulse rate, but a higher sampling rate is required for accurate analysis [ 28 ]. Therefore, data interpolation is performed with respect to the measured RPPG to be equal to the number of CPPG data at a sampling rate of 255.…”