Improved Heart Rate Estimation from Photoplethysmography During Physical Exercise Using Combination of NLMS and RLS Adaptive Filters

Abstract

Monitoring Health parameters like Heart Rate (HR), Blood Pressure (BP), Oxygen level in blood (SpO2) etc., is demanding in recent times that has led to many wearable devices. HR is estimated from Photoplethysmography (PPG) signal using PPG sensors. The HR estimation during physical exercise is challenging because of the presence of Motion Artifacts (MA). A novel de-noising algorithm that reduces MA present in PPG signal is proposed. In this proposed de-noising algorithm, the combination of Normalized Least Mean Square (NLMS) and Recursive Least Squares (RLS) adaptive filters are used. The PPG signal de-noised by NLMS and RLS adaptive filters are combined together by means of convex combination. The HR is estimated in the frequency domain by taking the Fast Fourier Transform (FFT) of the de-noised signal. The performance of this technique is evaluated using the IEEE Signal Processing cup 2015 dataset. The HR estimation is improved by using the proposed de-noising algorithm and an error of 0.96 beats per minute (BPM) is achieved.