A Robust Algorithm for Estimating Heart-Rate from Motion-Corrupted Photoplethysmographic Signals Using Adaptive Filtering and Nonparametric Spectrum Estimation

Abstract

Accurate heart rate monitoring from motion-corrupted photoplethysmographic (PPG) signals is very challenging due to the unpredicted nature of motion artifacts contaminating in the recorded signal. By availing reference artifact signals, adaptive filtering can be applied to the noisy signal to get a cleansed PPG signal from which the heart rate can be estimated. Three-axis acceleration data can be used as the reference signal for adaptive filtering. Some of the earlier methods use sparse signal decomposition, signal reconstruction, and spectrum estimation methods to estimate the heart rate. Here instead, a method of the least mean square adaptive filtering with decomposed acceleration signals as reference signals and spectrum estimation using periodogram is proposed. Since the reference signals used for adaptive filtering are decomposed into singular components the convergence of adaptive filter will not be hampered. Signal sparsification and reconstruction were not used and the whole processing was done on the exact recorded signals which increases the robustness of the method. Also, the simplicity of nonparametric spectrum estimation method leverages the system.