Wavelet based photoplethysmogram foot delineation for heart rate variability applications

A novel and easily implemented delineation algorithm is presented, which allows the foot of the photoplethsymogram to be located with sufficient accuracy for heart rate variability applications. This algorithm combines classical delineation techniques with the robustness of the wavelet transform. It can be implemented with a set of FIR filters and simple non-adaptive thresholding, making it suitable for real-time ambulatory applications. Results show that the accuracy of the algorithm matches that of a standard electrocardiogram delineation algorithm, the current standard for heart rate variability applications. The algorithm presented herein is also compared against four state-of-the-art delineation algorithms. Using a database that contains exercise data from thirteen patients across six activity levels and 7012 beats, a temporal accuracy of 3.8±2.6 ms (mean±std) was achieved with a sensitivity of 99.29% and a positive predictive value of 99.23%.